MOLECULAR SIGNATURE OF LIVER TUMOR GRADE AND USE TO EVALUATE PROGNOSIS AND THERAPEUTIC REGIMEN

- INSTITUT PASTEUR

The present invention concerns a method to determine the gene expression profile on a sample previously obtained from a patient diagnosed for a liver tumor, comprising assaying the expression of a set of genes in this sample and determining the gene expression profile (signature). In a particular embodiment, said method enables to determine the grade of liver tumor, such as hepatoblastoma (HB) or a hepatocellular carcinoma (HCC). The invention is also directed to kits comprising a plurality of pairs of primers or a plurality of probes specific for a set of genes, as well as to solid support or composition comprising a set of probes specific for a set of genes. These methods are useful to determine the grade of a liver tumor in sample obtained from a patient, to determine the risk of developing metastasis and/or to define the therapeutic regimen to apply to a patient.

Skip to: Description  ·  Claims  · Patent History  ·  Patent History
Description

The present invention relates to a method to in vitro determine the grade of a liver tumor in a sample previously obtained from a patient, using a molecular signature based on the expression of a set of genes comprising at least 2, especially has or consist of 2 to 16 genes, preferably a set of 16 genes. In a particular embodiment, the method focuses on hepatoblastoma (HB) or hepatocellular carcinoma (HCC), in adults or in children. The invention is also directed to sets of primers, sets of probes, compositions, kits or arrays, comprising primers or probes specific for a set of genes comprising at least 2 genes, especially has or consists of 2 to 16 genes, preferably exactly 16 genes. Said sets, kits and arrays are tools suitable to determine the grade of a liver tumor in a patient.

The liver is a common site of metastases from a variety of organs such as lung, breast, colon and rectum. However, liver is also a site of different kinds of cancerous tumors that start in the liver (primary liver cancers). The most frequent is the Hepatocellular Carcinoma (HCC) (about 3 out of 4 primary liver cancers are this type) and is mainly diagnosed in adults. In the United States approximately 10,000 new patients are diagnosed with hepatocellular carcinoma each year. Less frequent liver tumours are cholangiocarcinoma (CC) in adults and hepatoblastoma (HB) in children.

The prognosis and treatment options associated with these different kinds of cancers is difficult to predict, and is dependent in particular on the stage of the cancer (such as the size of the tumor, whether it affects part or all of the liver, has spread to other places in the body or its aggressiveness). Therefore, it is important for clinicians and physicians to establish a classification of primary liver cancers (HCC or HB) to propose the most appropriate treatment and adopt the most appropriate surgery strategy. Some factors are currently used (degree of local invasion, histological types of cancer with specific grading, tumour markers and general status of the patient) but have been found to not be accurate and sufficient enough to ensure a correct classification.

As far as the HB is concerned, the PRETEXT (pre-treatment extent of disease) system designed by the International Childhood Liver Tumor Strategy Group (SIOPEL) is a non invasive technique commonly used by clinicians, to assess the extent of liver cancer, to determine the time of surgery and to adapt the treatment protocol. This system is based on the division of the liver in four parts and the determination of the number of liver sections that are free of tumor (Aronson et al. 2005; Journal of Clinical Oncology; 23 (6): 1245-1252). A revised staging system taking into account other criteria, such as caudate lobe involvement, extrahepatic abdominal disease, tumor focality, tumor rupture or intraperitoneal haemorrhage, distant metastases, lymph node metastases, portal vein involvement and involvement of the IVC (inferior vena cava) and/or hepatic veins, has been recently proposed (Roebuck; 2007; Pediatr Radiol; 37: 123-132). However, the PRETEXT system, even if reproducible and providing good prognostic value, is based on imaging and clinical symptoms, making this system dependent upon the technicians and clinicians. There is thus a need for a system, complementary to the PRETEXT system, based on genetic and molecular features of the liver tumors.

The present invention concerns a method or process of profiling gene expression for a set of genes, in a sample previously obtained from a patient diagnosed for a liver tumor. In a particular embodiment said method is designed to determine the grade of a liver tumor in a patient.

By “liver tumor” or “hepatic tumor”, it is meant a tumor originating from the liver of a patient, which is a malignant tumor (comprising cancerous cells), as opposed to a benign tumor (non cancerous) which is explicitly excluded. Malignant liver tumors encompass two main kinds of tumors: hepatoblastoma (HB) or hepatocellular carcinoma (HCC). These two tumor types can be assayed for the presently reported molecular signature. However, the present method may also be used to assay malignant liver tumors which are classified as unspecified (non-HB, non-HCC).

The present method may be used to determine the grade of a liver tumor or several liver tumors of the same patient, depending on the extent of the liver cancer. For convenience, the expression “a liver tumor” will be used throughout the specification to possibly apply to “one or several liver tumor(s)”. The term “neoplasm” may also be used as a synonymous of “tumor”.

In a particular embodiment, the tumor whose grade has to be determined is located in the liver. The presence of the tumor(s) in the liver may be diagnosed by ultrasound scan, x-rays, blood test, CT scans (computerised tomography) and/or MRI scans (magnetic resonance imaging).

In a particular embodiment, the tumor, although originating from the liver, has extended to other tissues or has given rise to metastasis.

In a particular embodiment, the patient is a child i.e., a human host who is under 20 years of age according to the present application. Therefore, in a particular embodiment, the liver tumor is a paediatric HB or a paediatric HCC. In another embodiment, the liver tumor is an adult HCC.

A grade is defined as a subclass of the liver tumor, corresponding to prognostic factors, such as tumor status, liver function and general health status. The present method of the invention allows or at least contributes to differentiating liver tumors having a good prognosis from tumors with a bad prognosis, in terms of evolution of the patient's disease. A good prognosis tumor is defined as a tumor with good survival probability for the patient (more than 80% survival at two years for HB and more than 50% survival at two years for HCC), low probability of metastases and good response to treatment for the patient. In contrast, a bad prognosis tumor is defined as a tumor with an advanced stage, such as one having vascular invasion or/and extrahepatic metastasis, and associated with a low survival probability for the patient (less than 50% survival in two years).

The method of the invention is carried out on a sample isolated from the patient who has previously been diagnosed for the tumor(s) and who, optionally, may have been treated by surgery. In a preferred embodiment, the sample is the liver tumor (tumoral tissue) or of one of the liver tumors identified by diagnosis imaging and obtained by surgery or a biopsy of this tumor. The tumor located in the liver tumor is called the primary tumor.

In another embodiment, the sample is not the liver tumor, but is representative of this tumor. By “representative”, it is meant that the sample is regarded as having the same features as the primary tumors, when considering the gene expression profile assayed in the present invention. Therefore, the sample may also consist of metastatic cells (secondary tumors spread into different part(s) of the body) or of a biological fluid containing cancerous cells (such as blood).

The sample may be fixed, for example in formalin (formalin fixed). In addition or alternatively, the sample may be embedded in paraffin (paraffin-embedded) or equivalent products. In particular, the tested sample is a formalin-fixed, paraffin-embedded (FFPE) sample.

One advantage of the method of the present invention is that, despite the possible heterogeneity of some liver tumors (comprising epithelial tumor cells at different stages of liver differentiation within the same tumor), the assay has proved to be reproducible and efficient on liver tumor biopsies obtained from any part of the whole tumor. Therefore, there is no requirement for the isolation of cells presenting particular features except from the fact that they are obtained from a liver tumor or are representative thereof, to carry out the gene expression profile assay.

In a particular embodiment, the tumor originates from a patient having a Caucasian origin, in particular European, North American, Australian, New-Zealander or Afrikaners.

In a first step, the method or process of the invention comprises assaying the expression level of a set of genes in a sample, in order to get an expression profile thereof.

By “expression of a set of genes” (or “gene expression”), it is meant assaying, in particular detecting, the product or several products resulting from the expression of a gene, this product being in the form of a nucleic acid, especially RNA, mRNA, cDNA, polypeptide, protein or any other formats. In a particular embodiment, the assay of the gene expression profile comprises detecting a set of nucleotide targets, each nucleotide target corresponding to the expression product of a gene encompassed in the set.

The expression “nucleotide target” means a nucleic acid molecule whose expression must be measured, preferably quantitatively measured. By “expression measured”, it is meant that the expression product(s), in particular the transcription product(s) of a gene, are measured. By “quantitative” it is meant that the method is used to determine the quantity or the number of copies of the expression products, in particular the transcription products or nucleotide targets, originally present in the sample. This must be opposed to the qualitative measurement, whose aim is to determine the presence or absence of said expression product(s) only.

A nucleotide target is in particular a RNA, and most particularly a total RNA. In a preferred embodiment, the nucleotide target is mRNA or transcripts. According to the methods used to measure the gene expression level, the mRNA initially present in the sample may be used to obtain cDNA or cRNA, which is then detected and possibly measured.

In an embodiment, the expression of the gene is assayed directly on the sample, in particular in the tumor. In an alternative embodiment, the expression products or the nucleotide targets are prepared from the sample, in particular are isolated or even purified. When the nucleotide targets are mRNA, a further step comprising or consisting in the retro-transcription of said mRNA into cDNA (complementary DNA) may also be performed prior to the step of detecting expression. Optionally, the cDNA may also be transcribed in vitro to provide cRNA.

During the step of preparation, and before assaying the expression, the expression product(s) or the nucleotide target(s) may be labelled, with isotopic (such as radioactive) or non isotopic (such as fluorescent, coloured, luminescent, affinity, enzymatic, magnetic, thermal or electrical) markers or labels.

It is noteworthy that steps carried out for assaying the gene expression must not alter the qualitative or the quantitative expression (number of copies) of the expression product(s) or of the nucleotide target(s), or must not interfere with the subsequent step comprising assaying the qualitative or the quantitative expression of said expression product(s) or nucleotide target(s).

The step of profiling gene expression comprises determining the expression of a set of genes. Such a set is defined as a group of genes that must be assayed for one test, and especially performed at the same time, on the same patient's sample. A set comprises at least 2 and has especially from 2 to 16 genes, said 2 to 16 genes being chosen from the 16 following genes: alpha-fetoprotein (AFP), aldehyde dehydrogenase 2 (ALDH2), amyloid P component serum (APCS), apolipoprotein C-IV (APOC4), aquaporin 9 (AQP9), budding uninhibited by benzimidazoles 1 (BUB1), complement componant 1 (C1S), cytochrome p450 2E1 (CYP2E1), discs large homolog 7 (DLG7), dual specificity phosphatase 9 (DUSP9), E2F5 transcription factor (E2F5), growth hormone receptor (GHR), 4-hydroxyphenylpyruvase dioxygenase (HPD), immunoglogulin superfamily member 1 (IGSF1), Notchless homolog 1 (NLE1) and the ribosomal protein L10a (RPL10A) genes.

A complete description of these 16 genes is given in Table 1. This table lists, from left to right, the symbol of the gene, the complete name of the gene, the number of the SEQ ID provided in the sequence listing, the Accession Number from the NCBI database on June 2008, the human chromosomal location and the reported function (when known).

A set of genes comprises at least 2 out the 16 genes of Table 1, and particularly at least or exactly 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14 or 15 out of the 16 genes of Table 1. In a particular embodiment, the set comprises or consists of the 16 genes of Table 1 i.e. the set of genes comprises or consists of AFP, ALDH2, APCS, APOC4, AQP9, BUB1, C1S, CYP2E1, DLG7, DUSP9, E2F5, GHR, HPD, IGSF1, NLE1 and RPL10A genes. Accordingly, unless otherwise stated when reference is made in the present application to a set of 2 to 16 genes of Table 1, it should be understood as similarly applying to any number of genes within said 2 to 16 range.

In other particular embodiments, the set of genes comprises or consists of one of the following sets: (a) the E2F5 and HPD genes, (b) the APCS, BUB1, E2F5, GHR and HPD genes, (c) the ALDH2, APCS, APOC4, BUB1, C1S, CYP2E1, E2F5, GHR and HPD genes, (d) the ALDH2, APCS, APOC4, AQP9, BUB1, C1S, DUSP9, E2F5 and RPL10A genes, or (e) the ALDH2, APCS, APOC4, AQP9, C1S, CYP2E1, E2F5, GHR, IGSF1 and RPL10A genes.

As indicated by the expression “comprises from 2 to 16 genes of Table 1”, the set may, besides the specific genes of Table 1, contain additional genes not listed in Table 1. This means that the set must comprises from 2 to 16 genes of Table 1, i.e. 2 to 16 genes of Table 1 (in particular 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, or 16 genes), and optionally comprises one or more additional genes. Said set may also be restricted to said 2 to 16 genes of Table 1.

Additional genes may be selected for the difference of expression observed between the various grades of liver cancer, in particular between a tumor of good prognosis and a tumor of poor prognosis.

TABLE 1 mRNA Accession Protein symbol Gene name SEQ ID No Location Function SEQ ID AFP alpha-fetoprotein 1 NM_001134 4q11-q13 plasma protein synthesized 2 by the fetal liver ALDH2 aldehyde dehydrogenase 2 3 NM_000690 12q24.2 liver enzyme involved in 4 family (mitochondrial) alcohol metabolism APCS amyloid P component, serum 5 NM_001639 1q21-q23 secreted glycoprotein 6 AP0C4 apolipoprotein C-IV 7 NM_001646 19q13.2 secreted liver protein 8 AQP9 aquaporin 9 9 NM_020980 15q22.1-22.2 water-selective membrane channel 10 BUB1 BUB1 budding uninhibited 11 AF043294 2q14 kinase involved in spindle 12 by benzimidazoles 1 homolog checkpoint (yeast) C1S complement component 1, s 13 M18767 12p13 component of the cleavage and 14 subcomponent polyadenylation specificity factor complex CYP2E1 cytochrome P450, family 2, 15 AF182276 10q24.3-qter cytochrome P450 family member subfamily E, polypeptide 1 involved in drug metabolism DLG7 discs, large homolog7 17 NM_014750 14q22.3 cell cycle regulator involved 18 (Drosophila) (DLGAP5) in kinetocore formation DUSP9 dual specificity phosphatase 9 19 NM_001395 Xq28 phosphatase involved in 20 regulation of MAP Kinases E2F5 E2F transcription factor 5, 21 U15642 8q21.2 transcription factor involved in cell 22 p130-binding cycle regulation GHR Growth hormone receptor 23 NM_000163 5p13-p12 transmembrane receptor for 24 growth hormone HPD 4-hydroxyphenylpyruvate 25 NM_002150 12q24-qter enzyme involved in amino-acid 26 dioxygenase degradation IGSF1 immunoglobulin superfamily, 27 NM_001555 Xq25 cell recognition and 28 member 1 regulation of cell behavior NLE1 notchless homolog 1 29 NM_08096 17q12 unknown 30 (Drosophila) RPL10A ribosomal protein L10a 31 NM_007104 6p21.3-p21.2 ribosomal protein of 60S subunit 32

The invention also relates to a set of genes comprising or consisting of the 16 genes of Table 1 (i.e., AFP, ALDH2, APCS, APOC4, AQP9, BUB1, C1S, CYP2E1, DLG7, DUSP9, E2F5, GHR, HPD, IGSF1, NLE1 and RPL10A genes), in which 1, 2, 3, 4 or 5 genes out of the 16 genes are substituted by a gene presenting the same features in terms of difference of expression between a tumor of a good prognosis and a tumor of poor prognosis.

In a particular embodiment, the number of genes of the set does not exceed 100, particularly 50, 30, 20, more particularly 16 and even more particularly is maximum 5, 6, 7, 8, 9 or 10.

When considering adding or substituting a gene or several genes to the disclosed set, the person skilled in the art will consider one or several of the following features:

    • (a) the added gene(s) and/or the substituted gene(s) of Table 1 must present the same features in terms of difference of expression between a tumor of a good prognosis and a tumor of poor prognosis as the genes of Table 1 when taken as a whole. Thus, the expression of the added gene or of the substituted gene in a tumor of a good prognosis is either overexpressed or underexpressed of a factor of at least 2, preferably of at least 5, and more preferably of at least 10, as compared to its expression in a tumor of poor prognosis.
    • (b) besides presenting the feature in a), the added gene and/or the substituted gene may also provide, in combination with the other genes of the set, discriminant results with respect to the grade of the liver tumors; this discrimination is reflected by the homogeneity of expression profile of this gene in the tumors of a good prognosis on the one hand, and the tumors of poor prognosis in the other hand; and
    • (c) finally, besides features of a) and/or b), the added gene and/or the substituted gene is optionally chosen among genes that are involved in liver differentiation, in particular having a specific expression in fetal liver, or genes that are involved in proliferation, for example in mitosis or associated with ribosomes.

Examples of genes which can be added or may replace genes of the set may be identified in following Table 2.

TABLE 2 list of genes according to p value. Gene mean mean ratio Parametric symbol rC1 rC2 rC2/rC1 p-value FDR Description IPO4 123.7 248.3 2.0 2.00E−07 0.00036 importin 4 CPSF1 467.8 1010.7 2.2 2.00E-07 0.00036 cleavage and polyadenylation specific factor 1, 160 kDa MCM4 25.8 90.7 3.5 1.10E−06 0.00115 MCM4 minichromosome maintenance deficient 4 (S. cerevisise) EIF3S3 1319 2601.2 2.0 1.20E−06 0.00119 eukaryotic translation initiation factor 3, subunit 3 gamma, 40 kDa NCL 1319 2655.6 2.0 1.30E−06 0.00122 nucleolin CDC25C 35.7 99.3 2.8 1.40E−06 0.00124 cell division cycle 25C CENPA 28.2 78.4 2.8 1.50E−06 0.00124 centromere protein A, 17 kDa KIF14 24.7 54.2 2.2 1.50E−06 0.00124 kinesin family member 14 IPW 145.7 397.6 2.7 1.90E−06 0.0015 imprinted in Prader-Willi syndrome KNTC2 26.8 65.1 2.4 2.20E−06 0.00157 kinetochore associated 2 TMEM48 264 71.7 2.7 2.30E−06 0.00157 transmembrane protein 48 BOP1 87.2 270.9 3.1 2.30E−06 0.00157 block of proliferation 1 EIF3S9 170 372.4 2.2 2.30E−06 0.00157 eukaryotic translation initiation factor 3, subunit 9 eta, 116 kDa PH-4 340.9 168.2 0.5 2.40E−06 0.00158 hypoxia-inducible factor prolyl 4- hydroxylase SMC4L1 151.5 359.3 2.4 2.50E−06 0.0016 SMC4 structural maintenance of chromosomes 4-like 1 (yeast) TTK 23.7 74.2 3.1 2.60E−06 0.00161 TTK protein kinase LAMA3 696 136.3 0.2 2.80E−06 0.00168 laminin, alpha 3 C10orf72 192.6 67.7 0.4 2.90E−06 0.00169 Chromosome 10 open reading frame 72 TPX2 73.4 401.5 5.5 3.10E−06 0.00171 TPX2, microtubule-associated, homolog (Xenopus laevis) MSH2 75.5 212.1 2.8 3.20E−06 0.00171 mutS homolog 2, colon cancer, nonpolyposis type 1 (E. coli) DKC1 358.1 833.5 2.3 3.20E−06 0.00171 dyskeratosis congenita 1, dyskerin STK6 86.4 395.3 4.6 3.30E−06 0.00172 serine/threonine kinase 6 CCT6A 200.5 526.6 2.6 3.50E−06 0.00173 chaperonin containing TCP1, subunit 6A (zeta 1) SULT1C1 67.5 314.8 4.7 3.50E−06 0.00173 sulfotransferase family, cytosolic, 1C, member 1 ILF3 142.3 294.5 2.1 3.70E−06 0.00174 interleukin enhancer binding factor 3, 90 kDa IMPDH2 916.9 2385.6 2.6 3.70E−06 0.00174 IMP (inosine monophosphate) dehydrogenase 2 HIC2 63.4 208.8 3.3 3.90E−06 0.00179 hypermethylated in cancer 2 AFM 1310.3 237.4 0.2 4.10E−06 0.00184 afamin MCM7 187.3 465.3 2.5 4.30E−06 0.00189 MCM7 minichromosome maintenance deficient 7 (S. cerevisiae) CNAP1 70.2 177.5 2.5 4.40E−06 0.00189 chromosome condensation-related SMC- associated protein 1 CBARA1 958 475 0.5 4.60E−06 0.00194 calcium binding atopy-related autoantigen 1 PLA2G4C 123.3 51.2 0.4 4.90E−06 0.00194 phospholipase A2, group IVC (cytosolic, calcium-independent) CPSF1 301.9 616 2.0 5.00E−06 0.00194 cleavage and polyadenylation specific factor 1, 160 kDa SNRPN 30.9 100.6 3.3 5.00E−06 0.00194 Small nuclear ribonucleoprotein polypeptide N RPL5 2754.8 4961 1.8 5.20E−06 0.00194 ribosomal protein L5 C1R 1446.5 366.4 0.3 5.30E−06 0.00194 complement component 1, r subcomponent C16orf34 630.4 1109.6 1.8 5.30E−06 0.00194 chromosome 16 open reading frame 34 PHB 309.3 915.1 3.0 5.30E−06 0.00194 prohibitin BZW2 387.4 946.4 2.4 5.40E−06 0.00194 basic leucine zipper and W2 domains 2 ALAS1 1075.8 466.5 0.4 5.50E−06 0.00194 aminolevulinate, delta-, synthase 1 FLJ20364 48.6 112.4 2.3 5.70E−06 0.00198 hypothetical protein FLJ20364 RANBP1 593.7 1168.1 2.0 5.90E−06 0.00201 RAN binding protein 1 SKB1 354.7 687.4 1.9 6.20E−06 0.00208 SKB1 homolog (S. pombe) ABHD6 402.2 196.9 0.5 6.50E−06 0.00213 abhydrolase domain containing 6 CCNB1 60.4 330 5.5 6.60E−06 0.00213 cyclin B1 NOL5A 246.9 716.2 2.9 7.00E−06 0.00213 nucleolar protein 5A (56 kDa with KKE/D repeat) RPL8 3805.7 7390.5 1.9 7.00E−06 0.00213 ribosomal protein L8 BLNK 211.1 39.8 0.2 7.10E−06 0.00213 B-cell linker BYSL 167.3 269.7 1.6 7.10E−06 0.00213 bystin-like UBE1L 247.6 142.3 0.6 7.20E−06 0.00213 ubiquitin-activating enzyme E1-like CHD7 118.6 312 2.6 7.40E−06 0.00215 chromodomain helicase DNA binding protein 7 DKFZp762E1 70.2 219.4 3.1 7.60E−06 0.00218 hypothetical protein DKFZp762E1312 312 (HJURP) NUP210 178.4 284.9 1.8 7.70E−06 0.00218 nucleoporin 210 kDa PLK1 72.8 185.2 2.5 7.90E−06 0.0022 polo-like kinase 1 (Drosophila) ENPEP 116.2 29.4 0.3 8.00E−06 0.0022 glutamyl aminopeptidase (aminopeptidase A) HCAP-G 17.7 57.8 3.3 8.40E−06 0.00228 chromosome condensation protein G UGT2B4 1117.8 246.7 0.2 9.20E−06 0.00245 UDP glucuronosyltransferase 2 family, polypeptide B4 C20orf27 129.7 245.3 1.9 9.30E−06 0.00245 chromosome 20 open reading frame 27 C6orf149 178.7 491.1 2.7 9.40E−06 0.00245 chromosome 6 open reading frame 149 (LYRM4)
The Accession Numbers of the genes of Table 2. as found in NCBI database in June 2008, are the following: IPO4 (BC136759), CPSF1 (NM_013291), MCM4 (NM_005914.2; NM_182746.1; two accession numbers for the same gene correspond to 2 different isoforms of the gene), EIF3S3 (NM_003756.2), NCL (NM_005381.2), CDC25C (NM_001790.3), CENPA (NM_001809.3; NM_001042426.1), K1F14 (BC113742), IPW
# (U12897), KNTC2 (AK313184), TMEM48 (NM_018087), BOP1 (NM_015201), EIF3S9 (NM_003751; NM_001037283). PH-4 (NM _177939), SMC4L1 (NM_005496; NM_001002800), TTK (AK315696), LAMA3 (NM_198129), C10orf72 (NM_001031746; NM_144984), TPX2 (NM_012112), MSH2 (NM_000251), DKC1 (NM_001363), STK6 (AY892410), CCT6A (NM_001762; # NM_001009186), SULT1C1 (AK313193), ILF3 (NM_012218; NM_004516), IMPDH2 (NM_000884), HIC2 (NM_015094), AFM (NM_001133), MCM7 (NM_005916; NM_182776), CNAP1(AK128354), CBARA1 (AK225695), PLA2G4C (NM_003706), CPSF1(NM_013291), SNRPN (BC000611), RPL5 (AK314720), C1R (NM_001733), C16orf34 (CH471112), PHB (AK312649), BZW2 (BC017794), ALAS1(AK312566), # FLJ20364 (NM_017785), RANBP1 (NM_002882), SKB1 (AF015913), ABHD6 (NM_020676), CCNB1 (NM_031966), NOL5A (NM_006392), RPL8 (NM_000973; NM_033301), BLNK (NM_013314; NM_001114094), BYSL (NM_004053), UBE1L(AY889910), CHD7 (NM_017780), DKFZp762E1312 (NM_018410), NUP210(NM_024923), PLK1(NM_005030), ENPEP(NM_001977), # HCAP-G(NM_022346), UGT2B4 (NM_021139), C20orf27 (NM_001039140) and C6orf149 (NM_020408).

In a particular embodiment of the invention, the set of genes of the invention is designed to determine the grade of hepatoblastoma, in particular paediatric hepatoblastoma. In another embodiment, the set of genes is designed to determine the grade of hepatocellular carcinoma, in particular paediatric HCC or adult HCC.

The expression of the genes of the set may be assayed by any conventional methods, in particular any conventional methods known to measure the quantitative expression of RNA, preferably mRNA.

The expression may be measured after carrying out an amplification process, such as by PCR, quantitative PCR (qPCR) or real-time PCR. Kits designed for measuring expression after an amplification step are disclosed below.

The expression may be measured using hybridization method, especially with a step of hybridizing on a solid support, especially an array, a macroarray or a microarray or in other conditions especially in solution. Arrays and kits of the invention, designed for measuring expression by hybridization method are disclosed below.

The expression of a gene may be assayed in two manners:

    • to determine absolute gene expression that corresponds to the number of copies of the product of expression of a gene, in particular the number of copies of a nucleotide target, in the sample; and
    • to determine the relative expression that corresponds to the number of copies of the product of expression of a gene, in particular the number of copies of a nucleotide target, in the sample over the number of copies of the expression product or the number of copies of a nucleotide target of a different gene (calculation also known as normalisation). This different gene is not one of the genes contained in the set to be assayed. This different gene is assayed on the same sample and at the same time as the genes of the set to be assayed, and is called an invariant gene or a normalizer. The invariant gene is generally selected for the fact that its expression is steady whatever the sample to be tested. The expression “steady whatever the sample” means that the expression of an invariant gene does not vary significantly between a normal liver cell and the corresponding tumor cell in a same patient and/or between different liver tumor samples in a same patient. In the present specification, a gene is defined as invariant when its absolute expression does not vary in function of the grade of the liver tumors, in particular does not vary in function of the grade of the HB or HCC tumor, and/or does not vary between liver tumor and normal liver cells.

In the present invention, the expression which is assayed is preferably the relative expression of each gene, calculated with reference to at least one (preferably 1, 2, 3 or 4) invariant gene(s). Invariant genes, suitable to perform the invention, are genes whose expression is constant whatever the grade of the liver tumors, such as for example ACTG1, EFF1A1, PNN and RHOT2 genes, whose features are summarized in Table 3. In a particular embodiment preferred, the relative expression is calculated with respect to at least the RHOT2 gene or with respect to the RHOT2 gene.

In another advantageous embodiment, the relative expression is calculated with respect to at least the PNN gene or with respect to the PNN gene. It may be calculated with respect to the RHOT2 and PNN genes.

The calculation of the absolute expression or of the relative expression of each gene of the set and of each invariant gene being assayed with the same method from the same sample, preferably at the same time, enables to determine for each sample a gene expression profile.

TABLE 3 Features of invariant genes. ACTG1, EEF1A1, PNN and RHOT2 proteins are defined in SEQ ID NOs: 34, 36, 38 and 40 respectively. symbol Gene name SEQ 10* Accession No Location Function ACTG1 actin, gamma 1 33 NM_001614 17q25 cytoplasmic actin cytoskeleton in nonmuscle cells EEF1A1 eukaryotic translation 35 NM13 001 402 6q14.1 enzymatic delivery of elongation factor 1 aminoacyl tRNAs to alpha 1 the ribosome PNN pinin, desmosome 37 NM_002687 14q21.1 transcriptional associated protein corepressor, RNA splicing regulator RHOT2 ras homolog gene 39 NM_138769 16p13.3 Signaling by Rho family, member T2 GTPases, mitochondrial protein

An additional step of the method or process comprises the determination of the grade of said liver tumor, referring to the gene expression profile that has been assayed. In a particular embodiment of the invention, the method is designed to determine the grade of hepatoblastoma, in particular paediatric hepatoblastoma. In another embodiment, the method is designed to determine the grade of hepatocellular carcinoma, in particular paediatric HCC or adult HCC.

According to a particular embodiment of the invention, in the step of the method which is performed to determine the grade of the liver tumor, a gene expression profile or a signature (preferably obtained after normalization), which is thus specific for each sample, is compared to the gene expression profile of a reference sample or to the gene expression profiles of each sample of a collection of reference samples (individually tested) whose grade is known, so as to determine the grade of said liver tumor. This comparison step is carried out with at least one prediction algorithm. In a particular embodiment, the comparison step is carried out with 1, 2, 3, 4, 5 or 6 prediction algorithms chosen in the following prediction algorithms: Compound Covariate Predictor (CCP), Linear Discriminator Analysis (LDA), One Nearest Neighbor (1NN), Three Nearest Neighbor (3NN), Nearest Centroid (NC) and Support Vector Machine (SVM). These six algorithms are part of the “Biometric Research Branch (BRB) Tools” developed by the National Cancer Institut (NCI) and are available on http://linus.nci.nih.gov/BRB-ArrayTools.html. Equivalent algorithms may be used instead of or in addition to the above ones. Each algorithm classifies tumors within either of the two groups, defined as tumors with good prognosis (such as C1) or tumors with bad prognosis (such as C2); each group comprises the respective reference samples used for comparison, and one of these two groups also comprises the tumor to be classified.

Therefore, when 6 algorithms are used, the grade of a tumor sample may be assigned with certainty to the class of good prognosis or to the class of bad prognosis, when 5 or 6 of the above algorithms classified the tumor sample in the same group. In contrast, when less than 5 of the above algorithms classify a tumor sample in the same group, it provides an indication of the grade rather than a definite classification.

Reference samples which can be used for comparison with the gene expression profile of a tumor to be tested are one or several sample(s) representative for tumor with poor prognosis (such as C2), one or several sample(s) representative of tumor with good prognosis (such as C1), one or several sample(s) of a normal adult liver and/or one or several sample(s) of a fetal liver.

Table 4 lists the level of expression of each gene of Table 1 depending upon the status of the reference sample i.e., robust tumor with poor prognostic and robust tumor with good prognostic. Examples of methods to identify such robust tumors are provided in the examples. The present invention provides a new classification method in this respect, which is based on discretization of continuous values.

TABLE 4 Level of expression of the genes of Table 1, with respect to the status of the robust tumors Nucleotide Expression status in robust tumor target with poor prognosis with good prognosis AFP overexpressed underexpressed ALDH2 underexpressed overexpressed APCS underexpressed ovorexpressed APOC4 underexpressed overexpressed AQP9 underexpressed overexpressed BUB1 overexpressed underexpressed C1S underexpressed overexpressed CYP2E1 underexpressed overexpressed DLG7 overexpressed underexpressed DUSP9 overexpressed underexpressed E2F5 overexpressed underexpressed GHR underexpressed overexpressed HPD underexpressed overexpressed IGSF1 overexpressed underexpressed NLE1 overexpressed underexpressed RPL10A overexpressed underexpressed

Reference samples usually correspond to so-called “robust tumor” for which all the marker genes providing the signature are expressed (either under expressed or overexpressed) as expected i.e., in accordance with the results disclosed in Table 5, when tested in similar conditions, as disclosed in the examples hereafter.

A robust tumor having an overexpression of one or several gene(s) selected among ALDH2, APCS, APOC4, AQP9, C1S, CYP2E1, GHR and HPD genes (these genes belong to the so-called group of differentiation-related genes), and/or an underexpression of one or several gene(s) selected among AFP, BUB1, DLG7, DUSP9, E2F5, IGSF1, NLE1 and RPL10A genes (these genes belong to the so-called group of proliferation-related genes), is an indicator of a robust liver tumor, in particular of a hepatoblastoma, with a good prognosis. A robust tumor having an overexpression of one or several gene(s) selected among AFP, BUB1, DLG7, DUSP9, E2F5, IGSF1, NLE1 and RPL10A genes, and/or an underexpression of one or several gene(s) among ALDH2, APCS, APOC4, AQP9, C1S, CYP2E1, GHR and HPD genes, is an indicator of a robust liver tumor, in particular of a hepatoblastoma, with a poor prognosis. In the present application, a gene is said “underexpressed” when its expression is lower than the expression of the same gene in the other tumor grade, and a gene is said “overexpressed” when its expression is higher than the expression of the same gene in the other tumor grade.

In a particular embodiment, Table 5 provides the gene expression profiles of the 16 genes of Table 1 in 13 samples of hepatoblastoma (HB) including 8 samples that have been previously identified as rC1 subtype and 5 samples that have been previously identified as rC2 subtype. This Table can therefore be used for comparison, to determine the gene expression profile of a HB tumor to be classified, with the robust tumors disclosed (constituting reference samples), for a set of genes as defined in the present application. Said comparison involves using the classification algorithms which are disclosed herein, for both the selected reference samples and the assayed sample.

TABLE 5
Normalized qPCR data of 16 genes in 13 HB samples including 8 samples of the rC1 subtype and 5 samples of the rC2 subtype (in grey). The qpCR values have been obtained by measuring the expression of the 16 genes in 8 samples of the rC1 subtype and 5 samples of the rC2 subtype by the SYBR green method using the primers as disclosed in Table 6 below and in the conditions reported in the examples, and normalized by the ROTH2 gene (primers in Table 7).

The method of the present invention is also suitable to classify new tumor samples, and to use them as new reference samples. Therefore, the gene expression values of these new reference samples may be used in combination or in place of some of the values reported in Table 5.

In another embodiment of the invention, the step of determining the tumor grade comprises performing a method of discretization of continuous values of gene expression obtained on the set of genes the tested patients' samples. Discretization is generally defined as the process of transforming a continuous-valued variable into a discrete one by creating a set of contiguous intervals (or equivalently a set of cutpoints) that spans the range of the variable's values. Discretization has been disclosed for use in classification performance in Lustgarten J. L. et al, 2008.

The inventors have observed that discretization can be effective in determining liver tumor grade, especially for those tumors described in the present application, including Hepatoblastoma (HB) or Hepatocellular carcinoma (HCC).

The discretization method is especially disclosed in the examples where it is illustrated by using data obtained on tumor samples wherein these data are those obtained from profiling the 16 genes providing the large set of genes for expression profiling according to the invention. It is pointed out that the discretization method may however be carried out on a reduced number of profiled genes within this group of 16 genes, starting from a set consisting of 2 genes (or more genes) including one (or more) overexpressed proliferation-related genes chosen among AFP, BUB1, DLG7, DUSP9, E2F5, IGSF1, NLE1 and RPL10A and one down-regulated differentiation-related gene chosen among ALDH2, APCS, APOC4, AQP9, C1S, CYP2E1, GHR, HPD, said genes being thus classified as a result of gene profiles observed on robust tumors with poor prognosis (according to the classification in Table 4 above). In particular embodiments of the discretization method, the number of assayed gene for expression profiling is 2, 4, 6, 8, 10, 12, 14 or 16 and the same number of genes in each category (either the group of overexpressed proliferation-related genes or the group of downregulated differentiation-related gene) is used to perform the method.

The invention thus relates to a method enabling the determination of the tumor grade on a patient's sample, which comprises a classification of the tumor through discretization according to the following steps:

    • measuring the expression and especially the relative (normalized) expression of each gene in a set of genes defined as the signature of the tumor, for example by quantitative PCR thereby obtaining data as Ct or preferably Delta Ct, wherein said set of genes is divided in two groups, a first group consisting of the proliferation-related genes and a second group consisting of the differentiation-related genes (as disclosed above),
    • comparing the values measured for each gene, to a cut-off value determined for each gene of the set of genes, and assigning a discretized value to each of said measured values with respect to said cut-off value, said discretized value being advantageously a “1” or a “2” value assigned with respect to the cut-off value of the gene and optionally, if two cut-offs values are used for one gene, a further discretized value such as a “1.5” or another value between “1” or “2” may be assigned for the measured values which are intermediate between the cut-offs values,
    • determining the average of the discretized values for the genes, in each group of the set of genes,
    • determining the ratio of the average for the discretized values for the proliferation-related genes on the average for the discretized values for the differentiation-related genes, thereby obtaining a score for the sample,
    • comparing the obtained score for the sample with one or more sample cut-off(s), wherein each cut-off has been assessed for a selected percentile,
    • determining the tumor grade as C1 or C2, as a result of the classification of the sample with respect to said sample cut-off.

The above defined ratio of average values may be alternatively calculated as the ratio of the average for the discresized values for the differentiation-related genes on the average for the discretized values for the proliferation-related genes, to obtain a score. If this calculation made is adopted the cut-offs values are inversed, i.e., are calculated as 1/xxx.

In order to carry out the discretization method of the invention, the data obtained on the assayed genes for profiling a patient's sample are preferably normalized with respect to one or more invariant gene(s) of the present invention, in order to prevent detrimental impact on the results that may arise from possible inaccuracy in the quantification of initial nucleic acid, especially RNA, in the sample.

Normalization with respect to one invariant gene only, especially when said invariant gene is RHOT2 gene has proved to be relevant in the results obtained by the inventors. Similarly normalization with respect to PNN gene would be an advantageous possibility because the gene does also not vary in expression.

In order to design a discretization method for the determination of tumor grade of an individual sample of a patient, according to the invention, cut-offs values have to be determined to allow the determination of the tumor grade. The cut-offs values can be determined experimentally by carrying out the following steps on expression profiling results obtained on a determined number of tumor samples:

    • defining a cut-off (threshold value) for each gene in the set of genes designed for the signature, said cut-off corresponding to the value of the absolute or preferably relative (i.e. normalized) expression of said gene at a selected percentile and said percentile being selected for each of two groups of genes defined in the set of genes. In order to do so, the set of profiled genes comprises the same number of genes within each of the 2 groups of genes consisting of the group of overexpressed proliferation-related genes encompassing AFP, BUB1, DLG7, DUSP9, E2F5, IGSF1, NLE1 and RPL10A and the group of down-regulated differentiation-related gene encompassing ALDH2, APCS, APOC4, AQP9, C15, CYP2E1, GHR, HPD (said groups being defined based on gene profiles on robust tumors with poor prognosis),
    • in each tumor sample assigning to each expression value (especially normalized expression value) obtained for each expression profiled gene in the sample, a discretized value which is codified with respect to the cut-off value determined for the same gene and in line with the defined contiguous intervals of continuous values, e.g. a discretized value of “1” or “2” if two intervals (categories) are defined or a discretized value of “1”, “1.5” (or another value between 1 and 2) or “2” if three intervals are defined, said assignment of discretized value being advantageously such that the “1” is assigned for expression values falling below the cut-off found for the differentiation-related genes and for expression values falling below the cut-off found for the proliferation-related genes, the “2” is assigned for expression values falling above the cut-off found for the differentiation-related genes and for expression values falling above the cut-off found for the proliferation-related genes, and optionally if a “1.5” is used it is assigned to values found between the cut-offs;
    • on each tumor sample, determining in each group (proliferation-related genes group or differentiation-related genes group) the average value of said assigned discretized values of profiled genes of the set of profiled genes;
    • determining a score for each sample, as the ratio between the average expression values of said genes in said two groups of genes in the set of profiled genes;
    • determining on the basis of the obtained scores for all the tumor samples, one or more cut-off value(s) for the sample, corresponding to the respective value(s) at one or more (especially 2 or 3) percentile(s), wherein said percentile(s) is (are) either identical or different from the percentiles(s) selected for the genes.
      When the cut-offs values for each gene of the set of genes for profiling have been obtained for a sufficient number of relevant samples and the cut-off value for the sample is determined on the basis of the same samples, these cut-offs can be adopted as reference cut-offs for the user who will be carrying out the analysis of any further patient's tumor sample, especially for the purpose of determining the tumor grade in a patient's sample, if the analysis is performed in identical or similar conditions as the conditions which led to the establishment of the cut-offs values.

Therefore the invention provides cut-offs values as reference cut-offs, in order to carry out the determination of tumor grade in particular testing conditions as those disclosed below and in the examples.

In a particular embodiment of the method of discretization, the cut-off for each gene is the value corresponding to a determined percentile, which can be different for each of the considered two groups of genes (proliferation-related genes on the one hand and differentiation-related genes on the other hand). The selected percentile (or quantile) is determined with respect to the fraction of tumors (such as ⅓ or more) harbouring some chosen features such as overexpression of proliferation-related genes and/or dowregulation of differentiation-related genes, in the two groups of genes of the set of genes. Especially, when one intends to assign more weight to tumors displaying strong overexpression of proliferation-related genes and/or strong downregulation of differentiation-related genes, the cut-off corresponds to a high quantile (above the 50th, preferably the 60th, or even above the 65th, such as the 67th and for example within the range of 55th and 70th) for said proliferation-related genes and the cut-off corresponds to a low quantile (below the 50th, preferably equal to or below the 40th for example the 33rd, and for example within the range of between 20th and 40th) of the differentiation-related genes. The cut-off for each group of genes and the cut-off for the sample may be determined with respect to the same percentile(s) or may be determined with respect to different percentile.

According to a particular embodiment of the invention, for HB tumors, the percentile which is chosen for the overexpressed proliferation-related genes is the 67th and the percentile which is chosen for the downregulated differentiation-related genes is the 33rd. According to a particular embodiment of the invention, for HC tumors, the percentile which is chosen for the overexpressed proliferation-related genes is the 60th and the percentile which is chosen for the downregulated differentiation-related genes is the 40rd.

Each percentile (or cut-off value corresponding to the percentile) defines a cutpoint and the discretized values for each gene are either “1” or “2” below or above said percentile. The values “1” and “2” are distributed with respect to the percentiles so as to create the highest difference in the values of the calculated ratio for the most different tumor grades. This is illustrated in the examples for the selected percentiles.

It has been observed that in a preferred embodiment of the invention, the relative values of the profiled genes are determined by real-time PCR (qPCR).

Conditions to carry out the real-time PCR are disclosed herein, especially in the examples, as conditions applicable to analyzed samples.

PCR primers and probes suitable for the performance of RT-PCR are those disclosed herein for the various genes.

In a particular embodiment of the invention, the analysed tumor is a hepatoblastoma and its grade is determined by discretization as disclosed above and illustrated in the examples, taking into account that:

    • the set of assayed genes for profiling is constituted of the 16 genes disclosed;
    • the invariant gene (of reference) is RHOT2;
    • the cut-offs value for each gene based on −dCt (minus delta Ct) measures) are:
      AFP: 3.96139596; ALDH2: 4.3590482; APCS: 4.4691582; APOC4: 2.03068712; AQP9: 3.38391456; BUB1: −1.41294708; C1S: 4.24839464; CYP2E1: 6.70659644; DLG7: −3.3912188; DUSP9: 2.07022648; E2F5: −0.72728656; GHR: −0.1505569200; HPD: 2.27655628; IGSF1: 0.1075015200; NLE: −0.02343571999; RPL10A: 6.19723876.
    • the cut-off value for the sample is 0.91 (for the 67th) and optionally a further the cut-off value for the sample is 0.615 (for the 33rd). In such a case, a sample with a score above 0.91 is classified into the C2 class and a sample with a score below 0.91 is classified into the C1 class. The reference to the cut-off at 0.615 may be used to refine the results for values between both cut-offs.

In another embodiment of the invention, the tumor is an hepatocellular carcinoma and its grade is determined by discretization as disclosed above and illustrated in the examples, taking into account that:

    • the set of assayed genes for profiling is constituted of the 16 genes disclosed;
    • the invariant gene (of reference) is RHOT2;

the cut-offs value for each gene based on −dCt (minus delta Ct) measures) are:

Gene name Cut-off for Taqman Cut-off for SybrGreen AFP −1.2634010 −2.3753035 ALDH2 4.014143 5.314302 APCS 5.6142907 6.399079 APQC4 −0.7963158 4.656336 AQP9 4.2836011 5.446966 BUB1 −1.2736579 −3.634476 C1S 6.3514679 6.240002 CYP2E1 6.9562419 5.829384 DLG7 −2.335694 −4.614352 DUSP9 −7.979559 −1.8626715 E2F5 −0.4400218 −1.367846 GHR 1.0832632 1.169362 HPD 6.7480328 6.736329 IGSF1 −4.8417785 7.6653982 NLE −1.6167268 −1.82226 RPL10A 6.2483056 5.731897
    • the cut-off value for the score of a sample based on the ration between the average of the discretized values of the “proliferation-related genes” on the “differentiation-related genes” are 0.66 determined as the 30th percentile of the score) and 0.925 (determined as the 67th percentile of the score) In such a case, a sample with a score above 0.925 is classified into the C2 class and a sample with a score below 0.66 is classified into the C1 class. The sample with a score (initial score) between 0.66 and 0.925 can be assigned to an intermediate class. It can alternatively be classified as C1 or C2 using a modified score corresponding to the average of the discretized values of the “proliferation-related genes”. A new cut-off value is determined for said genes, which is the cut-off value for the modified score (in the present case it is 1.3). This cut-off can be determined via a percentile (here the 60th) of the distribution of the modified scores, using the samples of the intermediate class. A sample (initially classified in the intermediate class) with a modified score below 1.3 can be re-classified into the C1 class, and a sample with a modified score above 1.3 can be re-classified into the C2 class.

It is observed that the refinement of the results which are between the cut-offs of the samples is advantageous for hepatocellular carcinoma in order to increase the relevancy of the information on the tumor grade.

Generally said refinement of the classification of the intermediate results in the HCC is obtained by performing the following steps:

a modified score is determined which corresponds to the average of the discretized values of the “proliferation-related genes” only for the sample. A new cut-off value is determined for said genes, which is the cut-off value for the modified score (in the present case it is 1.3). This cut-off can be determined via a percentile (here the 60th) of the distribution of the modified scores, using the samples of the intermediate class. A sample (initially classified in the intermediate class) with a modified score below the “proliferation cut-off” (for example 1.3) can be re-classified into the C1 class, and a sample with a modified score above the “proliferation cut-off” (for example 1.3) can be re-classified into the C2 class.

From the 16 genes expressed in liver cells listed in Table 1, a set comprising from 2 to 16 genes (or more generally a set as defined herein) may be used to assay the grade of tumor cells in a tumor originating from the liver. The results obtained, after determining the expression of each of the genes of the set, are then treated for classification according to the steps disclosed herein. The invention relates to each and any combination of genes disclosed in Table 1, to provide a set comprising from 2 to 16 of these genes, in particular a set comprising or consisting of 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 or 16 of these genes. In the designed set, one or many genes of Table 1 may be modified by substitution or by addition of one or several genes as explained above, which also enable to determine the grade of the liver tumor, when assayed in combination with the other genes.

In a preferred embodiment, the liver tumor is a paediatric HB, and the method or process of the invention enables to distinguish a first class, called C1, qualifying as a good prognosis tumor and a second class, called C2, qualifying as a poor prognosis tumor. The C1 grade is predominantly composed of fetal histotype cells (i.e., well differentiated and non proliferative cells). In contrast, the C2 grade presents cells other than the fetal histotype such as embryonic, atypic (crowded fetal), small cell undifferiantiated (SCUD) and/or macrotrabecular cells.

The present invention also relates to a kit suitable to determine the grade of a liver tumor from the sample obtained from a patient. This kit is appropriate to carry out the method or process described in the present application.

In a particular embodiment, the kit comprises a plurality of pairs of primers specific for a set of genes to be assayed, said set comprising from 2 to 16 genes, said 2 to 16 genes being chosen in the group consisting of AFP, ALDH2, APCS, APOC4, AQP9, BUB1, C1S, CYP2E1, DLG7, DUSP9, E2F5, GHR, HPD, IGSF1, NLE1 and RPL10A genes.

By “plurality”, it is mean that the kit comprises at least as many pairs of primers as genes to enable assaying each selected gene, and in particular the nucleotide target of this gene. Accordingly, each gene and in particular its nucleotide target is specifically targeted by a least one of these pairs of primers. In a particular embodiment, the kit comprises the same number of pairs of primers as the number of genes to assay and each primer pair specifically targets one of the genes, and in particular the nucleotide targets of one of these genes, and does not hybridize with the other genes of the set.

The kits of the invention are defined to amplify the nucleotide targets of the sets of genes as described in the present invention. Therefore, the kit of the invention comprises from 2 to 16 pairs of primers which, when taken as a whole, are specific for said from 2 to 16 genes out of the 16 genes of Table 1. In particular, the kit comprises or consists of 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14 or 15 pairs of primers specific for 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14 or 15 out of the 16 genes of Table 1. In a particular embodiment, the kit comprises or consists of 16 pairs of primers specific for the 16 genes of Table 1 i.e., a primer pair specific for each of the following genes: AFP, ALDH2, APCS, APOC4, AQP9, BUB1, C1S, CYP2E1, DLG7, DUSP9, E2F5, GHR, HPD, IGSF1, NLE1 and RPL10A genes.

When the set of genes has been modified by the addition or substitution of at least one gene as described above, the kit is adapted to contain a pair of primers specific for each added or substituted gene(s). As indicated by the term “comprises”, the kit may, besides the pairs of primers specific for the genes of Table 1, contain additional pair(s) of primers.

In a particular embodiment, the kit comprises at least one pair of primers (preferably one) for at least one invariant gene (preferably one or two) to be assayed for the determination of the expression profile of the genes, by comparison with the expression profile of the invariant gene.

The number of pairs of primers of the kit usually does not exceed 100, particularly 50, 30, 20, more particularly 16, and even more particularly is maximum 5, 6, 7, 8, 9 or 10.

In the kits of the invention, it is understood that, for each gene, at least one pair of primers and preferably exactly one pair, enabling to amplify the nucleotide targets of this gene, is present. When the kits provide several pairs of primers for the same gene, the gene expression level is measured by amplification with only one pair of primers. It is excluded that amplification may be performed using simultaneously several pairs of primers for the same gene.

As defined herein, a pair of primers consists of a forward polynucleotide and a backward polynucleotide, having the capacity to match its nucleotide target and to amplify, when appropriate conditions and reagents are brought, a nucleotide sequence framed by their complementary sequence, in the sequence of their nucleotide target.

The pairs of primers present in the kits of the invention are specific for a gene i.e., each pair of primers amplifies the nucleotide targets of one and only one gene among the set. Therefore, it is excluded that a pair of primers specific for a gene amplifies, in a exponential or even in a linear way, the nucleotide targets of another gene and/or other nucleic acids contained in sample. In this way, the sequence of a primer (whose pair is specific for a gene) is selected to be not found in a sequence found in another gene, is not complementary to a sequence found in this another gene and/or is not able to hybridize in amplification conditions as defined in the present application with the sequence of the nucleotide targets of this another gene.

In a particular embodiment, the forward and/or backward primer(s) may be labelled, either by isotopic (such as radioactive) or non isotopic (such as fluorescent, biotin, fluororochrome) methods. The label of the primer(s) leads to the labelling of the amplicon (product of amplification), since the primers are incorporated in the final product.

The design of a pair of primers is well known in the art and in particular may be carried out by reference to Sambrook et al. (Molecular Cloning, A laboratory Manual, Third Edition; chapter 8 and in particular pages 8.13 to 8.16). Various softwares are available to design pairs of primers, such as Oligo™ or Primer3.

Therefore, each primer of the pair (forward and backward) has, independently from each other, the following features:

    • their size is from 10 and 50 bp, preferably 15 to 30 bp; and
    • they have the capacity to hybridize with the sequence of the nucleotide targets of a gene.

In a particular embodiment, when the pairs of primers are used in a simultaneous amplification reaction carried out on the sample, the various primers have the capacity to hybridize with their respective nucleotide targets at the same temperature and in the same conditions.

Conventional conditions for PCR amplification are well known in the art and in particular in Sambrook et al. An example of common conditions for amplification by PCR is dNTP (200 mM), MgCl2 (0.5-3 mM) and primers (100-200 nM).

In a particular embodiment, the sequence of the primer is 100% identical to one of the strands of the sequence of the nucleotide target to which it must hybridize with, i.e. is 100% complementary to the sequence of the nucleotide target to which it must hybridize. In another embodiment, the identity or complementarity is not 100%, but the similarity is at least 80%, at least 85%, at least 90% or at least 95% with its complementary sequence in the nucleotide target. In a particular embodiment, the primer differs from its counterpart in the sequence of the sequence of the nucleotide target by 1, 2, 3, 4 or 5 mutation(s) (deletion, insertion and/or substitution), preferably by 1, 2, 3, 4 or 5 nucleotide substitutions. In a particular embodiment, the mutations are not located in the last 5 nucleotides of the 3′ end of the primer.

In a particular embodiment, the primer, which is not 100% identical or complementary, keeps the capacity to hybridize with the sequence of the nucleotide target, similarly to the primer that is 100% identical or 100% complementary with the sequence of the nucleotide target (in the hybridization conditions defined herein). In order to be specific, at least one of the primers (having at least 80% similarity as defined above) of the pair specific for a gene can not hybridize with the sequence found in the nucleotide targets of another gene of the set and of another gene of the sample.

In a particular embodiment, the pairs of primers used for amplifying a particular set of genes are designed, besides some or all of the features explained herein, in order that the amplification products (or amplicons) of each gene have approximately the same size. By “approximately” is meant that the difference of size between the longest amplicon and the shortest amplicon of the set is less than 30% (of the size of the longest amplicon), preferably less than 20%, more preferably less than 10%. As particular embodiments, the size of the amplicon is between 100 and 300 bp, such as about 100, 150, 200, 250 or 300 bp.

The nucleotide sequences of the 16 genes of Table 1 are provided in the Figures, and may be used to design specific pairs of primers for amplification, in view of the explanations above.

Examples of primers that may be used to measure the expression of the genes of Table 1, in particular to amplify the nucleotide targets of the genes of Table 1, are the primers having the sequence provided in Table 6 or variant primers having at least 80% similarity (or more as defined above) with the sequences defined in Table 6.

TABLE 6 Sequence of forward and backward primers of the 16 genes defined in Table 1. These primers may be used in any real- time PCR, in particular the SYBR green technique, except for the Taqman ® protocol. Product size Target (bp) Forward primer (5′-3′) Reverse primer (5′-3′) AFP 151 AACTATTGGCCTGTGGCGAG TCATCCACCACCAAGCTGC ALDH2 151 GTTTGGAGCCCAGTCACCCT GGGAGGAAGCTTGCATGATTC APCS 151 GGCCAGGAATATGAACAAGCC CTTCTCCAGCGGTGTGATCA APOC4 151 GGAGCTGCTGGAGACAGTGG TTTGGATTCGAGGAACCAGG AQP9 151 GCTTCCTCCCTGGGACTGA CAACCAAAGGGCCCACTACA BUB1 152 ACCCCTGAAAAAGTGATGCCT TCATCCTGTTCCAAAAATCCG C1S 141 TTGTTTGGTTCTGTCATCCGC TGGAACACATTTCGGCAGC CYP2E1 151 CAACCAAGAATTTCCTGATCCAG AAGAAACAACTCCATGCGAGC DLG7 151 GCAGGAAGAATGTGCTGAAACA TCCAAGTCTTTGAGAAGGGCC DUSP9 151 CGGAGGCCATTGAGTTCATT ACCAGGTCATAGGCATCGTTG E2F5 151 CCATTCAGGCACCTTCTGGT ACGGGCTTAGATGAACTCGACT GHR 151 CTTGGCACTGGCAGGATCA AGGTGAACGGCACTTGGTG HPD 151 ATCTTCACCAAACCGGTGCA CCATGTTGGTGAGGTTACCCC IGSF1 152 CACTCACACTGAAAAACGCCC GGGTGGAGCAATTGAAAGTCA NLE1 151 ATGTGAAGGCCCAGAAGCTG GAGAACTTCGGGCCGTCTC RPL10A 151 TATCCCCCACATGGACATCG TGCCTTATTTAAACCTGGGCC

The kit of the invention may further comprise one or many pairs of primers specific for one or many invariant genes, in particular specific for ACTG1, EFF1A1, PNN and/or RHOT2 genes. The pair of primers specific for invariant gene(s) may be designed and selected as explained above for the pair of primers specific for the genes of the set of the invention. In a particular embodiment, the pairs of primers of the invariant genes are designed in order that their amplification product (or amplicon) has approximately the same size as the amplicon of the genes of the set to be assayed (the term approximately being defined as above, with respect to the longest amplicon of the set of genes). Examples of primers that may be used to amplify the particular invariant genes are primers having the sequence provided in Table 7 or primers having at least 80% similarity (or more as defined above) with the sequences defined in Table 7.

TABLE 7 Sequence of forward and backward primers specific for the invariant genes defined in Table 3. These primers may be used in real-time PCR, in particular the SYBR green technique, except for the Taqman ® protocol. Product size Target (bp) Forward primer (5′-3′) Reverse primer (5′-3′) ACTG1 151 GATGGCCAGGTCATCACCAT ACAGGTCTTTGCGGATGTCC EFF1A1 151 TCACCCGTAAGGATGGCAAT CGGCCAACAGGAACAGTACC PNN 151 CCTTTCTGGTCCTGGTGGAG TGATTCTCTTCTGGTCCGACG RHOT2 151 CTGCGGACTATCTCTCCCCTC AAAAGGCTTTGCAGCTCCAC

The kits of the invention may also further comprise, in association with or independently of the pairs of primers specific for the invariant gene(s), reagents necessary for the amplification of the nucleotide targets of the sets of the invention and if any, of the nucleotide targets of the invariant genes.

The kits of the invention may also comprise probes as disclosed herein in the context of sets of probes, compositions and arrays. In particular, the kits also comprise the four dNTPs (nucleotides), amplification buffer, a polymerase (in particular a DNA polymerase, and more particularly a thermostable DNA polymerase) and/or salts necessary for the activity of the polymerase (such as Mg2+).

Finally, the kits may also comprise one or several control sample(s) i.e., at least one sample(s) representative of tumor with bad (i.e., poor) prognosis (in particular a HB C2 grade), at least one sample(s) representative of tumor with good prognosis (in particular a HB C1 grade), at least one sample of a normal adult liver and/or at least one sample of a fetal liver.

The kits may also comprise instructions to carry out the amplification step or the various steps of the method of the invention.

The invention is also directed to a set of probes suitable to determine the grade of a liver tumor from the sample obtained from a patient. This set of probes is appropriate to carry out the method or process described in the present invention. It may also be part of the kit.

This set of probes comprises a plurality of probes in particular from 2 to 16 probes, these 2 to 16 probes being specific for genes chosen in the group consisting of AFP, ALDH2, APCS, APOC4, AQP9, BUB1, C1S, CYP2E1, DLG7, DUSP9, E2F5, GHR, HPD, IGSF1, NLE1 and RPL10A genes.

By “plurality”, it is mean that the set of probes comprises at least as many probes as genes to assay. In a particular embodiment, the array comprises the same number of probes as the number of genes to assay.

The probes of the sets of the invention are selected for their capacity to hybridize to the nucleotide targets of the sets of genes as described in the present invention. Therefore, the set of probes of the invention comprise from 2 to 16 probes specific for 2 to 16 genes out of the 16 genes of Table 1. In particular, the sets of probes comprise or consist of 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14 or 15 probes specific of 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14 or 15 out of the 16 genes of Table 1. In a particular embodiment, the sets of probes comprise or consist of 16 probes specific for the 16 genes of Table 1 i.e., a probe specific of each of the following genes: AFP, ALDH2, APCS, APOC4, AQP9, BUB1, C1S, CYP2E1, DLG7, DUSP9, E2F5, GHR, HPD, IGSF1, NLE1 and RPL10A genes.

The specificity of the probes is defined according to the same parameters as those applying to define specific primers.

When the set of genes has been modified by the addition or substitution of at least one gene as described above, the set of probes is adapted to contain a probe specific for the added or substituted gene(s). As indicated by the term “comprises”, the set of probes may, besides the probes specific for the genes of Table 1, contain additional probe(s).

The number of probes of the set does usually not exceed 100, particularly 50, 30, 20, more particularly 16, and even more particularly is maximum 5, 6, 7, 8, 9 or 10.

In the set of probes of the invention, it is understood that for each gene corresponds at least one probe to which the nucleotide target of this gene hybridize to. The set of probes may comprise several probes for the same gene, either probes having the same sequence or probes having different sequences.

As defined herein, a probe is a polynucleotide, especially DNA, having the capacity to hybridize to the nucleotide target of a gene. Hybridization is usually carried out at a temperature ranging from 40 to 60° C. in hybridization buffer (see example of buffers below). These probes may be oligonucleotides, PCR products or cDNA vectors or purified inserts. The size of each probe is independently to each other from 15 and 1000 bp, preferably 100 to 500 bp or 15 to 500 bp, more preferably 50 to 200 bp or 15 to 100 bp. The design of probes is well known in the art and in particular may be carried out by reference to Sambrook et al. (Molecular Cloning, A laboratory Manual, Third Edition; chapters 9 and 10 and in particular pages 10.1 to 10.10).

The probes may be optionally labelled, either by isotopic (radioactive) or non isotopic (biotin, fluororochrome) methods. Methods to label probes are disclosed in Sambrook et al. (Molecular Cloning, A laboratory Manual, Third Edition; chapter 8 and in particular page 9.3). In a particular embodiment, the probes are modified to confer them different physicochemical properties (such as by methylation, ethylation). In another particular embodiment, the probes may be modified to add a functional group (such as a thiol group), and optionally immobilized on bead (preferably glass beads).

In a particular embodiment, the sequence of the probe is 100% identical to a part of one strand of the sequence of the nucleotide target to which it must hybridize, i.e. is 100% complementary to a part of the sequence of the nucleotide target to which it must hybridize. In another embodiment, the identity or complementarity is not 100% and the similarity is at least 80%, at least 85%, at least 90% or at least 95% with a part of the sequence of the nucleotide target. In a particular embodiment, the probe differs from a part of one strand of the sequence of the nucleotide target by 1 to 10 mutation(s) (deletion, insertion and/or substitution), preferably by 1 to 10 nucleotide substitutions. By “a part of”, it is meant consecutive nucleotides of the nucleotide target, which correspond to the sequence of the probe.

In a particular embodiment, the probe, which is not 100% identical or complementary, keeps the capacity to hybridize, in particular to specifically hybridize, to the sequence of the nucleotide target, similarly to the probe which is 100% identical or 100% complementary with the sequence of the nucleotide target (in the hybridization conditions defined herein).

In a particular embodiment, the size of the probes used to assay a set of genes is approximately the same for all the probes. By “approximately” is meant that the difference of size between the longest probe and the shortest probe of the set is less than 30% (of the size of the longest probe), preferably less than 20%, more preferably less than 10%.

The set of probes of the invention may further comprise at least one (preferably one) probe specific for at least one invariant gene (preferably one or two), in particular specific for ACTG1, EFF1A1, PNN and/or RHOT2 genes. The probes specific for invariant gene(s) may be designed and selected as explained above for the probes specific for genes of the sets of the invention. In a particular embodiment, the probes specific of the invariant genes have approximately the same size as the probes specific of the genes of the set of be assayed (the term approximately being defined as above, with respect to the longest probes of the set of genes).

The invention is also directed to an array suitable to determine the grade of a liver tumor from the sample obtained from a patient. This array is appropriate to carry out the method or process described in the present application.

An array is defined as a solid support on which probes as defined above, are spotted or immobilized. The solid support may be porous or non-porous, and is usually glass slides, silica, nitrocellulose, acrylamide or nylon membranes or filters.

The arrays of the invention comprise a plurality of probes specific for a set of genes to be assayed. In particular, the array comprises, spotted on it, a set of probes as defined above.

The invention also relates to a composition comprising a set of probes as defined above in solution.

In a first embodiment, the probes (as defined above in the set of probes) may be modified to confer them different physicochemical properties (such as methylation, ethylation). The nucleotide targets (as defined herein and prepared from the sample) are linked to particles, preferably magnetic particles, for example covered with ITO (indium tin oxide) or polyimide. The solution of probes is then put in contact with the target nucleotides linked to the particles. The probe/target complexes are then detected, for example by mass spectrometry.

Alternatively, probes may be modified to add a functional group (such as a thiol group) and immobilized on beads (preferably glass beads). These probes immobilized on beads are put in contact with a sample comprising the nucleotide targets, and the probe/target complexes are detected, for example by capillary reaction.

The invention is also directed to kits comprising the sets of probes, the compositions or the arrays of the invention and preferably the primer pairs disclosed herein. These kits may also further comprise reagents necessary for the hybridization of the nucleotide targets of the sets of genes and/or of the invariant genes, to the probes (as such, in the compositions or on the arrays) and the washing of the array to remove unbound nucleotides targets.

In a particular embodiment, the kits also comprise reagents necessary for the hybridization, such as prehybridization buffer (for example containing 5×SSC, 0.1% SDS and 1% bovine serum albumin), hybridization buffer (for example containing 50% formamide, 10×SSC, and 0.2% SDS), low-stringency wash buffer (for example containing 1×SSC and 0.2% SDS) and/or high-stringency wash buffer (for example containing 0.1×SSC and 0.2% SDS).

The kits may also comprise one or several control sample(s) i.e., at least one sample(s) representative for tumor with poor prognosis, at least one sample(s) representative of tumor with good prognosis, at least one sample of a normal adult liver and/or at least one sample of a fetal liver. Alternatively, it may comprise the representation of a gene expression profile of such tumors.

Finally, the invention provides a kit as described above further comprising instructions to carry out the method or process of the invention.

The arrays and/or kits (either comprising pairs of primers or probes or arrays or compositions of the invention or all the components) according to the invention may be used in various aspects, in particular to determine the grade of a liver tumor from a patient, especially by the method disclosed in the present application.

The arrays and/or kits according to the invention are also useful to determine, depending upon the grade of the liver tumor, the risk for a patient to develop metastasis. Indeed, the classification of a liver tumor in the class with poor prognosis is highly associated with the risk of developing metastasis.

In another embodiment, the arrays and/or kits according to the invention are also useful to define, depending upon the grade of the liver tumor, the therapeutic regimen to apply to the patient.

The invention also relates to a support comprising the data identifying the gene expression profile obtained when carrying out the method of the invention.

BRIEF DESCRIPTION OF THE DRAWINGS

The colour version of the drawings as filed is available upon request to the European Patent Office.

FIG. 1. Identification of Two HB Subclasses by Expression Profiling.

(A) Schematic overview of the approach used to identify robust clusters of samples, including two tumor clusters (rC1 and rC2) and one non-tumor cluster (NL) (B) Expression profiles of 982 probe sets (824 genes) that discriminate rC1 and rC2 samples (p<0.001, two-sample t test). Data are plotted as a heatmap where red and green correspond to high and low expression in log2-transformed scale. (C) Molecular classification of 25 HB samples and status of CTNNB1 gene and β-catenin protein. C1 and C2 classification was based on rC1 and rC2 gene signature by using six different statistical predictive methods (CCP, LDA, 1NN, 3NN, NC and SVM) and the leave-one-out cross-validation. Black and gray squares indicate mutations of the CTNNB1 and AXIN1 genes. Immunohistochemical analysis of β-catenin in representative C1 and C2 cases is shown. (D) Expression of representative Wnt-related and β-catenin target genes (p<0.005, two-sample t test) in HB subclasses and non-tumor livers (NL). (E) Classification of hepatoblastoma by expression profile of a 16-gene signature. (F) Classification of normal human livers of children with HB (from 3 months to 6 years of age) (NT) or fetal livers at 17 to 35 weeks of gestation (FL) by expression profile of a 16-gene signature.

FIG. 2: Molecular HB subclasses are related to liver development stages. (A) Distinctive histologic and immunostaining patterns of HB subclasses C1 and C2. From top to bottom: numbers indicate the ratio of mixed epithelial-mesenchymal tumors and of tumors with predominant fetal histotype in C1 and C2 subtypes; hematoxylin and eosin (H&E) and immunostaining of Ki-67, AFP and GLUL in representative samples. Magnification, ×400. (B) Expression of selected markers of mature hepatocytes and hepatoblast/liver progenitors in HB subclasses and non-tumor livers.

FIG. 3: Validation of the 16-gene signature by qPCR in an independent set of 41 HBs. Expression profiles of the 16 genes forming the HB classifier are shown as a heatmap that indicates high (red) and low (green) expression according to log2-transformed scale. HB tumors, HB biopsies (b) and human fetal livers (FL) at different weeks (w) of gestation were assigned to class 1 or 2 by using the 16-gene expression profile, six different statistical predictive methods (CCP, LDA, 1NN, 3NN, NC and SVM) and leave-one-out cross-validation. Black boxes in the rows indicate from top to bottom: human fetal liver, mixed epithelial-mesenchymal histology, predominant fetal histotype, and β-catenin mutation.

FIG. 4: Gene expression of the 16 genes of the prognostic liver cancer signature assessed by qPCR is presented as box-plot. The boxes represent the 25-75 percentile range, the lines the 10-90 percentile range, and the horizontal bars the median values.

FIG. 5: Expression level of the 16 liver prognostic signature genes shown case by case in 46 hepatoblastomas and 8 normal livers. C1 tumors (green), C2 tumors (red) and normal liver (white).

FIG. 6. Correlation between molecular HB subtypes and clinical outcome in 61 patients. (A) Association of clinical and pathological data with HB classification in the complete set of 61 patients. Only significant correlations (Chi-square test) are shown. PRETEXT IV stage indicates tumorous involvement of all liver sections. (B) Kaplan-Meier plots of overall survival for 48 patients that received preoperative chemotherapy. Profiling via the 16-gene expression signature was used to define C1 and C2 subclasses in tumors resected after chemotherapy, and differences between survival curves were assessed with the log-rank test. (C) Overall survival of 17 HB patients for which pretreatment biopsies or primary surgery specimens were available. The signature was applied exclusively to tumor samples without prior therapy. (D) Multivariate analysis including 3 variables associated to patient's survival. The predominant histotype is defined as either fetal or other (including embryonal, crowed-fetal, macrotrabecular or SCUD types). Tumor stage is defined by PRETEXT stage (Perilongo et al., 2000) and/or distant metastasis at diagnosis and/or vascular invasion. HR, Hazard Ratio; CI, Confidence Interval.

FIG. 7: Clinical, pathological and genetic characteristics of 61 HB cases. SR: standard risk; HR: high risk according to SIOPEL criteria; NA: not available; PRETEXT: pre-treatment extent of disease according to SIOPEL; DOD: dead of disease; *: Vascular invasion was defined by radiological analysis; **: The predominant epithelial histotype variable categorized as “others” included embryonal, crowded fetal, macrotrabecular, and undifferentiated histotypes.

FIG. 8: Clinical, pathological and genetic characteristics of 66 HB samples; Tumor ID number indicates patient number. When more than one sample from the same patient was analyzed, the representative sample used for statistical analysis of clinical correlations is marked by an asterisk; b: biopsy. HB74F: fetal component of HB74; HB74e: embryonal component of HB74. Gender: M, male; F, female; Y, yes; N, no; NA, not available. Multifocality: S, solitary nodules; M, multiple nodules. Histology: E, epithelial; M, mixed; CF, crowded fetal; F, fetal; E, embryonal; M, macrotrabecular; PF, pure fetal; S, SCUD. PRETEXT β-catenin status: wt, wild-type; Δex3, in-frame deletion of part or all exon 3 sequence; FAP, familial polyposis kindred; AXIN1, Axin 1 nonsense mutation (R533stop, CGA to TGA).stage: I to IV according to SIOPEL (Aronson et al., 2005). Treatment protocol: S, standard risk; H, high risk according to SIOPEL. Outcome: A, alive free of disease; DOD, dead of disease; D, death unrelated to cancer; R, alive with recurrence of disease.

FIG. 9: Correlation between molecular HB subtypes and clinical outcome in 86 patients. (A) Association of clinical and pathological data with HB classification in the complete set of 86 patients. Only significant correlations (Chi-square test) are shown. PRETEXT IV stage indicates tumorous involvement of all liver sections. (B) Kaplan-Meier plots of overall survival for 73 patients that received preoperative chemotherapy. Profiling via the 16-gene expression signature was used to define C1 and C2 subclasses in tumors resected after chemotherapy, and differences between survival curves were assessed with the log-rank test. (C) Overall survival of 29 HB patients for which pretreatment biopsies or primary surgery specimens were available. The signature was applied exclusively to tumor samples without prior therapy. (D) Multivariate analysis including 3 variables associated to patient's survival. The predominant histotype is defined as either fetal or other (including embryonal, crowed-fetal, macrotrabecular or SCUD types). Tumor stage is defined by PRETEXT stage (Perilongo et al., 2000) and/or distant metastasis at diagnosis and/or vascular invasion. HR, Hazard Ratio; CI, Confidence Interval.

FIG. 10: Correlation between molecular HCC subtypes and clinical outcome in 64 patients. Kaplan-Meier estimates of overall survival in 64 HCC patients using molecular classification with 16 genes, with the unsupervised clustering (centroid) (A) or unsupervised clustering (average) (B).

FIG. 11: Analysis of the probability of overall survival (OS) of 85 hepatoblastoma patients using Kaplan-Meier estimates. Left pannel: cases were classified by the discretization method into 3 classes using as cut-offs the 33rd percentile and the 67th percentile. Middle pannel: cases were classified into 2 classes using the 33rd percentile. Right pannel: cases were classified into 2 classes using the 67th percentile.

FIG. 12: Analysis of the probability of overall survival (OS) or disease-free survival (DFS) of 113* HCC patients using Kaplan-Meier estimates and log-rank test.
Among the total series of 114 patients, survival data were not available for one case.

Patients were treated either by partial hepatectomy (PH) or by orthotopic liver transplantation (OLT). Unless specified, the follow-up was closed at 146 months.

A: HCC cases were classified into 3 classes by the discretization method using as cut-offs the 33rd and the 67th percentiles.

B: 47 HCC cases previously classified into the intermediate class (33<p<67, see pannel A) were subdivided into 2 new subclasses using the 60th percentile of proliferation-related genes.

C: 92 HCC cases treated by partial hepatectomy (PH) were classified into 3 classes as in pannel A.

D: 21 HCC cases treated by orthotopic liver transplantation (OLT) were classified into 2 classes using as cut-off the 67th percentile.

E: HCC cases were classified into 2 classes using different combinations of scores as described in Table F.

F: HCC cases were classified into 2 classes using as cut-off the 33rd percentile.

G: HCC cases were classified into 2 classes using as cut-off the 50th percentile.

H: HCC cases were classified into 2 classes using as cut-off the 67th percentile.

I: 92 HCC cases treated by partial hepatectomy (PH) were classified into 2 classes using as cut-off the 33rd percentile.

J: 92 HCC cases treated by partial hepatectomy (PH) were classified into 2 classes using as cut-off the 50th percentile.

K: 92 HCC cases treated by partial hepatectomy (PH) were classified into 2 classes using as cut-off the 67th percentile.

L: Disease-free survival of 113 HCC cases after classification into 2 classes using as cut-off the 67th percentile. Follow-up was closed at 48 months. Data were not significant when the follow-up was closed at 146 months.

M: Disease-free survival of 92 HCC cases treated by PH, after classification into 2 classes using as cut-off the 67th percentile. Follow-up was closed at 48 months. Data were not significant when the follow-up was closed at 146 months.

FIG. 13: Analysis of the probability of overall survival (OS) or disease-free survival (DFS) HCC patients using Kaplan-Meier estimates and log-rank test.

EXAMPLES Experimental Procedures

A. Patients and Tissue Samples.

Sixty-six tumor specimens and biopsies from 61 patients with hepatoblastoma were collected from different hospitals in France (52 cases), Italy (6 cases), United Kingdom (1 case), Switzerland (1 case) and Slovakia (1 case). Forty-eight patients received chemotherapy treatment prior to surgery, most being enrolled in clinical trials of the International Childhood Liver Tumour Strategy Group (SIOPEL) (Perilongo et al., 2000). Samples from fresh tumors avoiding fibrotic and necrotic areas and from adjacent non tumor livers were snap frozen at the time of surgery and stored at −80° C. FIG. 7 describes patient characteristics and clinicopathological parameters.

Patients were children with median age of 2 years, and male:female ratio of 1.5. The median follow-up was 32 months; during this period, 15 patients died from disease. The histology of all tumor specimens was centrally reviewed by expert pathologist according to previously described criteria (Finegold et al., 2007; Zimmermann, 2005). Twenty-five tumors were analyzed on oligonucleotide microarrays and 24 of them, for which DNA was available, were subjected to aCGH analysis, while a second set of 41 tumors was analyzed by qPCR (FIG. 8). No difference was observed in significant clinical and pathological data as well as in the percentage of cases carrying β-catenin mutation between the two sets. This study has been approved by the Ethics Committee of Institut Pasteur, and informed consent of the families was obtained at each Medical Center, in accordance with European Guidelines for biomedical research and with national laws in each country.

B. Oligonucleotide Microarrays and Gene Expression Data Analysis

Twenty-five HB samples and 4 non-tumor samples including a pool of livers from 3 males and a second from 3 females were analyzed using Affymetrix HG-U133A oligonucleotide arrays. Total RNA was prepared using FastPrep® system (Qbiogene, Strasbourg, France) and RNeasy mini Kit (Qiagen, Courtaboeuf, France). RNA quality was checked with the Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, Calif.). Microarray experiments were performed according to the manufacturer's instructions. Affymetrix microarray data were normalized using RMA method (Irizarry et al., 2003). Class discovery was done as described elsewhere (Lamant et al., 2007). Pathway and Gene Ontology enrichment analyses were performed using GSEA method (Subramanian et al., 2005) and hypergeometric tests. For supervised tests and class prediction, we used Biometric Research Branch (BRB) ArrayTools v3.2.2 software, developed by R. Simon and A. Peng. Permutations of the measurements are then used to estimate the FDR (the percentage of genes identified by chance). Additionally, mouse fetal livers at E18.5 and postnatal livers at 8 days of birth were profiled on Affymetrix MG-U74A, B v2 arrays. Data were processed and analyzed as aforementioned.

Except when indicated, transcriptome analysis was carried out using either an assortment of R system software packages (http://www.R-project.org, v2.3.0) including those of Bioconductor v1.8 (Gentleman et al., 2004) or original R code.

B.1. Normalization

Raw data from Affymetrix HG-U133A 2.0 GeneChip™ microarrays were normalized in batch using robust multi-array average method (R package affy, v1.10.0) (Irizarry et al., 2003). Probe sets corresponding to control genes or having a “_x_” annotation were masked yielding a total of 19,787 probe sets available for further analyses.

B.2. Class Discovery

Step 1

Variance Test

The variance of each probe set across samples was tested and compared to the median variance of all the probe sets, using the model: ((n−1)×Var(probe set)/Varmed), where n refers to the number of samples. By using the same filtering tool of BRB ArrayTools software, the P-value for each probe set was obtained by comparison of this model to a percentile of Chi-square distribution with (n−1) degrees of freedom.

Robust Coefficient of Variation (rCV)

The rCV was calculated for each probe set as follows. After ordering the intensity values of n samples from min to max, we eliminated the min and max values and we calculated the coefficient of variation (CV) for the remaining values.

Unsupervised Probe Sets Selection

Unsupervised selection of probe set lists was based on the two following criteria:

(i) variance test at P<0.01,

(ii) rCV less than 10 and superior to a given rCV percentile. We used eight rCV percentile thresholds (60%; 70%; 80%; 90%; 95%; 97.5%; 99%; 99.5%), which yielded 8 probe set lists.

Step 2: Generation of a Series of 24 Dendrograms

Hierarchical clustering was performed by using the 8 rCV-ranked probe sets lists, 3 different linkage methods (average, complete and Ward's), and 1-Pearson correlation as a distance metric (package cluster v1.9.3). This analysis generated 24 dendrograms.

Step 3:

Stability Assessment

The intrinsic stability of each of the 24 dendrograms was assessed by comparing each dendrogram to the dendrograms obtained after data “perturbation” or “resampling” (100 iterations). Perturbation stands for the addition of random gaussian noise (μ=0, σ=1.5×median variance calculated from the data set) to the data matrix, and resampling for the random substitution of 5% of the samples by virtual sample's profiles, generated randomly. The comparison between dendrograms across all iterations yielded a mean ‘similarity score’ (see below). The overall stability was assessed by calculating a mean similarity score, using all pairs of the 24 dendrograms.

Similarity Score

To compare two dendrograms, we compared the two partitions in k clusters (k=2 to 8) obtained from these two dendrograms. To compare a pair of partitions, we used a similarity measure, which corresponds to the symmetric difference distance (Robinson and Foulds, 1981).

Step 4: Identification of Robust Clusters

We identified groups in which any pair of samples was co-classified in at least 22 of the 24 partitions, and considered only groups made of 4 samples or more. Then, for any pair of these groups, we calculated the mean number of co-classification of any sample in the first group with any sample in the second group. We aggregated the groups for which this score was at least 18 (over the 24 partitions).

B.3. Supervised Tests

We compared gene expression between two classes of samples by using the Student's t test with random variance model option (BRB ArrayTools software, version 3.4.0a, developed by Dr. Richard Simon and Amy Peng Lam, http://linus.nci.nih.gov/BRB-ArrayTools.html). False Discovery Rates were assessed by using 1000 random permutations of labels (Monte Carlo approach).

B.4. Classification

To classify samples according to gene expression profile, we used the Class prediction tool of BRB ArrayTools software using all 6 following algorithms: Compound Covariate Predictor (CCP), Linear Discriminant Analysis (LDA), 1-Nearest Neighbor (1NN), 3-Nearest Neighbors (3NN), Nearest Centroid (NC) and Support Vector Machines (SVM). Each sample was classified according to the majority of the 6 algorithms. Samples classified as C2 by at least 3 algorithms were classified accordingly.

B.5. Gene Ontology and Pathway Analysis

We used a hypergeometric test to measure the association between a gene (probe set) list and a gene ontology term (GO term), as in GO stats R package (R. Gentleman). To this end, we mapped the gene list and the GO terms to non-redundant Entrez Gene identifiers by using the annotation file HG-U133_Plus—2.annot.csv (http://www.affymetrix.com, Dec. 14, 2006). GO terms and their relationships (parent/child) were downloaded from http://www.geneontology.org (version Dec. 31, 2006). The list of proteins associated to GO terms (table gene_association.goa_human) and mapping the Entrez Gene ids (table human.xrefs) were downloaded from ftp://ftp.ebi.ac.uk/pub/databases/GO/goa.

KEGG pathway annotation was done by Onto-tools software (http://vortex.cs.wayne.edu/ontoexpress/servlet/UserInfo). We designated a significance threshold of each hypergeometric test at P<0.001, and the condition that a GO term or pathway be represented by at least 3 Entrez Gene identifiers.

B.6. Gene Set Enrichment Analysis (gsea)

GSEA (Subramanian et al., 2005) was used to evaluate the correlation of a specific gene list with two different sample groups (phenotypes). Briefly, this method calculates an enrichment score after ranking all genes in the dataset based on their correlation with a chosen phenotype and identifying the rank positions of all the members of a defined gene set. We used the signal2noise ratio as a statistic to compare specific and random phenotypes in order to evaluate statistical differences.

C. Array-Based Comparative Genomic Hybridization (aCGH)

Genomic DNA from 24 HBs and 3 non-tumor liver samples was analyzed using aCGH chips designed by the CIT-CGH consortium. This array contains 3400 sequence-verified PAC/BAC clones spaced at approximately 1 Mb intervals, spotted in triplicate on Ultra Gaps slides (Corning Inc, Corning, N.Y.).

The aCGH chip was designed by CIT-CGH consortium (Olivier Delattre laboratory, Curie Institute, Paris; Charles Theillet laboratory, CRLC Val d'Aurelle, Montpellier; Stanislas du Manoir laboratory, IGBMC, Strasbourg and the company IntegraGen™). DNAs were labeled by the random priming method (Bioprime DNA labelling system; Invitrogen, Cergy-Pontoise, France) with cyanine-5 (Perkin-Elmer, Wellesley, Mass.). Using the same procedure, we labeled control DNAs with cyanine-3. After ethanol-precipitation with 210 μg of Human Cot-1 DNA (Invitrogen), resuspension in hybridization buffer (50% formamide), denaturation at 95° C. for 10 minutes and prehybridization at 37° C. for 90 minutes, probes were cohybridized on aCGH. The aCGH slides were previously preblocked with a buffer containing 2.6 mg succinic anhydride/118 ml N-methyl-2-pyrrolidinone/32 ml sodium tetraborate decahydrate, pH 8.0 (Sigma-Aldrich, Lyon, France). After washing, arrays were scanned using a 4000B scan (Axon, Union City, Calif.). Image analysis was performed with Genepix 5.1 software (Axon) and ratios of Cy5/Cy3 signals were determined. The aCGH data were normalized using lowess per block method (Dudoit et al., 2002). Comparison between groups was done using chi-square test or Fisher's exact test, as appropriate.

Status assignment (Gain/Loss) was performed using R package GLAD v1.6.0. Computation of recurrent minimal genomic alterations was done using slight modification of a previously described method (Rouveirol et al., 2006). For comparison between groups, we used the Fischer exact test. Complete aCGH data will be published elsewhere.

D. Mouse Microarray Analysis

Murine Genome Affymetrix U74v2 A and B arrays were used to investigate liver expression at embryonic day 18.5 (E18.5) and at 8 days after birth (PN8). Each time point consisted of a pool of livers from 3-5 animals analyzed in triplicate. Microarray experiments were performed according to the manufacturer's instructions.

Publicly available Affymetrix Mouse Genome (MG) 430 2.0 array liver expression data at embryonic time points E11.5, E12.5, E13.5, E14.5, and E16.5 days of gestation (Otu et al., 2007), were downloaded from the Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE6998).

MG-U74v2, MG-430 2.0 and HG-133A 2.0 array intra- and cross-species probeset comparison was achieved by using the Affymetrix NetAffx analysis center and by choosing “Good Match” degree of specificity. Unification of sample replicates, multiple array data standardization and Heatmap visualization was done by using dCHIP v1.6 software. Comparison of fetal liver stages by supervised analysis was performed using BRB ArrayTools software as previously described, by classing E11.5 and E12.5 as “Early” and E14.5 and E16.5 as “Late” fetal liver stage. Supervised signature was applied to HB array data, and intensity cut-off=60 was chosen in order to remove probesets that did not reach such intensity level in at least one sample.

E. Quantitative PCR Analysis (qPCR)

For qPCR analysis, we used RNA from 52 tumor samples (including 11 samples analyzed on microarrays, see FIG. 8), and from 8 non-tumor livers and 5 human fetal livers (RNAs purchased from BioChain Institute, Hayward, Calif.).

RNA was extracted by using either Trizol, RNeasy kit (QIAGEN) or miRvana kit (Ambion), then quantified and quality-checked by Agilent technology. For each cDNA preparation, 1 μg of RNA was diluted at the final concentration of 100 ng/μl, and reverse transcribed with the Superscript RT kit (Invitrogen, Carlsbad, Calif.) following the manufacturer's protocol. Random primers (Promega, Charbonnières-les-Bains, France) were added at the final concentration of 30 ng/μl and the final volume was 20 μl.

The cDNA was diluted 1:25, and 5 μl were used for each qPCR reaction. We added 5 μl of 2×Sybr Green Master mix (Applied Biosystems) and 0.3 μl of each specific primer (final concentration 300 nM). Each reaction was performed in triplicate. qPCR reactions were run on the Applied Biosystems 7900HT Fast Real-Time PCR System with a 384-well thermo-block, in the following conditions: 2 min at 50° C. to activate Uracil-N-glycosylase (UNG)-mediated erase of aspecific reaction; 10 min at 95° C. to activate the polymerase and inactivate the UNG; 40 cycles (15 sec at 95° C. denaturation step and 1 min at 60° C. annealing and extension); and final dissociation step to verify amplicon specificity.

The lists of primers used for qPCR are provided in Table 6 and Table 7 above.

F. Immunohistochemistry (IHC)

IHC was carried out as reported previously (Wei et al., 2000). For antigen retrieval at 95° C., we used 1 mM EDTA (pH 8) for β-catenin and Ki-67 IHC, and 10 mM citrate buffer (pH 6) for AFP and GLUL IHC. We used monoclonal antibodies against β-catenin and GLUL (Cat. Nos. 610154 and 610517; BD Biosciences, Le Pont de Claix, France) and Ki-67 (M7240, Dako, Trappes, France) and polyclonal antibody against AFP (N1501, Dako). Reactions were visualized using the ChemMate Dako Envision Detection kit (Dako) and diaminobenzidine. Subcellular distribution and quantitative evaluation of immunostaining in the different histotypes were assessed by examining at least ten random high-power fields.

G. Clinical Data Analysis

We used the Chi-square test for comparisons between groups. Survival curves were calculated according to the Kaplan-Meier method, using the log-rank test to assess differences between curves. Variables independently related to survival were determined by stepwise forward Cox regression analysis. Follow-up was closed at February 2007 or at time of death. Statistical analysis was done with SPSS software v10.0 (SPSS Inc., Chicago, Ill.).

H. Examples of Other Pairs of Primers and Probes for the 16 Genes of Table 1 and the 4 Invariant Genes (Table 3) that can be Used in the Taqman® Method.

AFP forward primer: GCCAGTGCTGCACTTCTTCA AFP reverse primer: TGTTTCATCCACCACCAAGCT AFP probe: ATGCCAACAGGAGGCCATGCTTCA (for each polynucleotide, the sequence is given from 5′ to 3′) ALDH2 forward primer: TGCAGGATGGCATGACCAT ALDH2 reverse primer: TCTTGAACTTCAGGATCTGCATCA ALDH2 probe: CCAAGGAGGAGATCTTCGGGCCA APCS forward primer: AGCTGGGAGTCCTCATCAGGTA APCS reverse primer: CGCAGACCCTTTTTCACCAA APCS probe: TGCTGAATTTTGGATCAATGGGACACC APOC4 forward primer: TGAAGGAGCTGCTGGAGACA APOC4 reverse primer: CGGGCTCCAGAACCATTG APOC4 probe: TGGTGAACAGGACCAGAGACGGGTG AQP9 forward primer: GCCATCGGCCTCCTGATTA AQP9 reverse primer: GTTCATGGCACAGCCACTGT AQP9 probe: TGTCATTGCTTCCTCCCTGGGACTG BUB1 forward primer: ACATCTGGTTTTCAGTGTGTTGAGA BUB1 reverse primer: GTTGCAGCAACCCCAAAGTAA BUB1 probe: TCAGCAACAAACCATGGAACTACCA GATCG C1S forward primer: TCCCAATGACAAGACCAAATTCT C1S reverse primer: AGAGCCCATAGGTCCCACACT C1S probe: CGCAGCTGGCCTGGTGTCCTG CYP2E1 forward CATGAGATTCAGCGGTTCATCA primer: CYP2E1 reverse GGTGTCTCGGGTTGCTTCA primer: CYP2E1 probe: CCTCGTGCCCTCCAACCTGCC DLG7 forward primer: GCTGGAGAGGAGACATCAAGAAC DLG7 reverse primer: CCTGGTTGTAGAGGTGAAAAAGTAATC DLG7 probe: TGCCAGACACATTTCTTTTGGTGGTAA CC DUSP9 forward primer: GGCCTACCTCATGCAGAAGCT DUSP9 reverse primer: GGGAGATGTTAGACTTCTTCCTCTTG DUSP9 probe: CACCTCTCTCTCAACGATGCCTATGA CCTG E2F5 forward primer: CCTGTTCCCCCACCTGATG E2F5 reverse primer: TTTCTGTGGAGTCACTGGAGTCA E2F5 probe: CCTCACACAGCCTTCCTCCCAGTCC GHR forward primer: CCCAGGTGAGCGACATTACA GHR reverse primer: CATCCCTGCCTTATTCTTTTGG GHR probe: CAGCAGGTAGTGTGGTCCTTTCCCCG HPD forward primer: CCCACGCTCTTCCTGGAA HPD reverse primer: TTGCCGGCTCCAAAACC HPD probe: TCATCCAGGGCCACAACCACCA IGSF1 forward primer: GACCATTGCCCTTGAAGAGTGT IGSF1 reverse primer: GAGAGGTTGATGAAGGAGAATTGG IGSF1 probe: ACCAAGAAGGAGAACCAGGCACCCC NLE1 forward primer: TGCCTCCTTTGACAAGTCCAT NLE1 reverse primer: CGCGTAGGGAAGCCAGGTA NLE1 probe: TGGGATGGGAGGAGGGGCA RPL10A forward primer TCGGCCCAGGTTTAAATAAGG RPL10A reverse primer CCACTTTGGCCACCATGTTT RPL10A Taqman probe AGTTCCCTTCCCTGCTCACACACAACG ACTG1 forward primer: GGCGCCCAGCACCAT ACTG1 reverse primer: CCGATCCACACCGAGTACTTG ACTG1 probe: ATCAAGATCATCGCACCCCCAGAGG EEF1A1 forward GCGGTGGGTGTCATCAAAG primer: EEF1A11 reverse TGGGCAGACTTGGTGACCTT primer: EEF1A11 probe: AGTGGACAAGAAGGCTGCTGGAGCTG PNN forward primer: GAATTCCCGGTCCGACAGA PNN reverse primer: TTTCGGTCTCTTTCACTTCTTGAA PNN probe: AGAGGTCTATATCAGAGAGTAGTCGA TCAGGCAAAAGA RHQT2 forward primer: CCCAGCACCACGATCTTCAC RHOT2 reverse primer: CCAGAAGGAAGAGGGATGCA RHOT2 Taqman probe: CAGCTCGCCACCATGGCCG

Results

Identification of Two HB Subclasses by Gene Expression Profiling

For robust unsupervised classification, we generated and screened a series of 24 dendrograms to identify samples that co-clustered whatever the method and the gene list. We obtained two robust subgroups of tumors named robust Cluster 1 (rC1, n=8) and robust Cluster 2 (rC2, n=5) (FIG. 1A). Comparison of rC1 and rC2 expression profiles identified 824 genes (p<0.001, false discovery rate (FDR)=0.02) (FIG. 1B). KEGG pathway analysis pinpointed a strong enrichment of cell cycle related genes (p<10−11), most being up-regulated in rC2 tumors. These: genes were mainly assigned to GO categories including mitosis regulation, spindle checkpoint, nucleotide biosynthesis, RNA helicase activity, ribosome biogenesis, and translational regulation. Evidence that rC2 tumors were faster proliferating than rC1 tumors was further confirmed by Ki-67 immunostaining (see FIG. 2A).

The remaining tumors were classified into C1 (rC1-related) and C2 (rC2-related) subclasses by applying a predictive approach based on the rC1/rC2 gene signature and using robust samples as training set (FIG. 1C). Both groups exhibited similar, high rates of β-catenin mutations, and accordingly, immunohistochemistry (IHC) of β-catenin showed cytoplasmic and nuclear staining of the protein in the majority of HBs. However, β-catenin localization was predominantly membranous and cytoplasmic in C1 tumors, whereas it showed frequent loss of membrane anchoring and intense nuclear accumulation in C2 tumors (FIG. 1C).

We observed differential expression of a number of Wnt members and targets between subclasses. C2 tumors showed increased expression of MYCN, BIRC5 that encodes the anti-apoptotic factor Survivin, NPM1 (encoding nucleophosmin) and HDAC2. By contrast, most C1 tumors prominently expressed the Wnt antagonist DKK3, BMP4, and genes previously found to be activated in liver tumors carrying mutant β-catenin (Boyault et al., 2007; Renard et al., 2007; Stahl et al., 2005). Remarkably, most genes related to liver functions are expressed in the perivenous area of adult livers, such as GLUL, RHBG, and two members of the cytochrome p450 family: CYP2E1 and CYP1A1 (Benhamouche et al., 2006; Braeuning et al., 2006) (FIG. 1D).

Further evidence that the rC1 subclass was enriched in genes assigned to the hepatic perivenous program was provided by Gene Set Enrichment Analysis (GSEA), a computational method for assessing enrichment of a predefined gene list in one class as compared with another (Subramanian et al., 2005). Thus, Wnt/β-catenin signaling appears to activate different transcriptional programs in HB subtypes, likely reflecting different cellular contexts.

HB Subclasses Evoke Distinct Phases of Liver Development

Next, we sought to determine whether HB subclasses were associated with specific histological phenotypes. Mixed epithelial-mesenchymal tumors that represented 20% of cases were not significantly associated with C1 and C2 subclasses. By contrast, a tight association was found with the main epithelial component, which defines the cell type occupying more than 50% of tumor cross-sectional areas. Sixteen out of 18 C1 tumors displayed a predominant fetal phenotype, including 4 ‘pure fetal’ cases, whereas all C2 tumors showed a more immature pattern, with prevailing embryonal or crowded-fetal histotypes associated with high proliferation (Finegold, 1994) (p<0.0001) (FIG. 2A). Further relationship between molecular subclasses and hepatic developmental stages was provided by the finding that a number of mature hepatocyte markers were markedly downregulated in C2 compared to C1 tumors (Tables 1 and 2). Conversely, C2 tumors showed strong overexpression (35-fold) of the oncofetal AFP gene associated to high protein levels in tumor cells by IHC (FIG. 2A) and in patients' sera (r=0.79, p<0.0001). C2 tumors also abundantly expressed hepatic progenitor markers such as KRT19 (encoding cytokeratin 19) and TACSTD1, also known as Ep-CAM (FIG. 2B).

To better define the relationships between HB subclasses and phases of hepatic differentiation, we first generated a liver development-related gene signature by making use of publicly available mouse fetal and adult liver data sets (Otu et al., 2007). When applied to HB samples, this signature was able to distinguish by hierarchical clustering two HB groups closely matching the C1/C2 classification. Next, we integrated HB gene expression data with the orthologous genes expressed in mouse livers at embryonic days (E) 11.5 to 18.5, and at 8 days of birth. In unsupervised clustering, most C2 tumors co-clustered with mouse livers at early stages of embryonic development (E11.5 and E12.5), whereas C1 tumors gathered with mouse livers at late fetal and postnatal stages. Together, these data comfort the notion that tumor cells in C2 and C1 subtypes are arrested at different points of the hepatic differentiation program.

Identification of a 16-Gene Signature as HB Classifier

To investigate the relevance of molecular HB classification in an independent set of tumors, we defined a HB classifier signature derived from the top list of genes differentially expressed between rC1 and rC2 clusters. After qPCR assessment, a list of 16 top genes at p≦10−7 was selected to form a class predictor (Table 1). Most of these genes show drastic variations in expression level during liver development, and among them, BUB1 and DLG7 have been repeatedly identified as hESC markers (Assou et al., 2007). The 16-gene expression profile was first investigated in rC1 and rC2 samples used as training set, and it predicted classification with 100% of accuracy in these samples, using either microarray or qPCR data. The robustness of this signature was confirmed by correct classification into C1 and C2 subclasses of all 13 remaining tumors analyzed by microarray (FIG. 1E). Expression profiles of fetal livers and normal liver for these 16-gene signature were also assayed (FIG. 1F). This signature was therefore employed to classify a new, independent set of 41 HB samples by qPCR (FIGS. 4 and 5 and Table 8), resulting in 21 tumors categorized as C1 and 20 tumors as C2 subtype (FIG. 3).

Extending our previous observation, C1/C2 classification in this new set of tumors was unrelated to CTNNB1 mutation rate. Using qPCR, we also confirmed enhanced expression in C2 tumors of liver progenitor markers such as AFP, Ep-CAM, and KRT19, as well as MYCN (FIG. 3). Moreover, while a similar percentage of C1 and C2 tumors displayed mesenchymal components, a predominant fetal histotype was found in 95% of tumors of the C1 subtype, whereas in 82% of C2 tumors, the major component displayed less differentiated patterns such as embryonal, crowded-fetal, macrotrabecular and SCUD types (p<0.0001) (FIG. 3). To further assess the association of HB subclasses with liver development, 5 human fetal livers at different weeks of gestation were included in the qPCR studies. In unsupervised clustering, fetal livers at late (>35 weeks) and earlier (17 to 26 weeks) developmental stages were classified as C1 and C2 respectively, further supporting that HB subclasses reflect maturation arrest at different developmental phases.

TABLE 8 Gene expression of the prognostic signature for liver cancer by quantitative RT-PCR. C1 C2 NL Fold-change median min max median min max median min max C1/NL C2/NL C2/C1 C1/C2 AFP 0.4 0.0 33.3 30.7 0.0 456.1 0.2 0.0 8.8 2.3 38.1 16.5 0.1 ALDH2 87.1 13.2 356.7 15.0 2.2 74.4 240.4 151.6 387.6 0.3 0.1 0.2 5.2 APCS 61.6 1.1 338.9 1.9 0.0 276.2 158.6 92.7 509.5 0.2 0.0 0.1 19.8 APOC4 21.3 4.3 122.8 1.6 0.1 24.2 47.0 22.3 112.4 0.5 0.0 0.1 16.1 AQP9 60.6 8.0 540.6 2.5 0.1 90.1 46.6 38.0 72.7 1.3 0.1 0.1 18.9 BUB1 0.0 0.0 0.4 0.9 0.1 3.9 0.0 0.0 0.1 1.2 16.1 13.4 0.1 C1S 51.1 14.9 277.2 7.5 1.3 96.0 223.4 129.3 565.3 0.2 0.0 0.2 5.7 CYP2E1 583.2 97.7 3463.0 19.7 0.4 1504.0 1128.6 527.6 1697.0 0.7 0.0 0.0 51.6 DLG7 0.0 0.0 0.0 0.1 0.0 0.5 0.0 0.0 0.0 1.7 12.4 7.3 0.1 DUSP9 1.5 0.4 45.7 19.1 0.0 179.0 0.6 0.2 1.3 4.0 18.3 4.6 0.2 E2F5 0.2 0.0 2.0 1.1 0.1 11.7 0.1 0.0 0.5 1.8 6.5 3.5 0.3 GHR 5.2 0.0 54.0 0.5 0.0 2.4 35.2 20.8 54.5 0.1 0.0 0.1 8.6 HPD 22.9 0.9 182.0 1.2 0.1 23.8 111.5 62.6 165.7 0.2 0.0 0.1 14.0 IGSF1 0.1 0.0 1.7 1.7 0.0 19.8 0.1 0.0 0.1 2.2 22.4 10.2 0.1 NLE 0.4 0.1 4.8 0.8 0.3 5.1 0.4 0.2 0.8 1.2 2.2 1.8 0.5 RPL10A 73.3 12.0 230.4 98.2 11.9 432.8 86.9 54.1 159.9 0.8 1.1 1.5 0.7
NL, non-tumor liver; C1, good prognosis hepatoblastomas; C2, bad prognosis hepatoblastomas. Shown are the median values of 46 hepatoblastomas from 41 patients, the minimal and maximal values in each class, and the fold changes between classes. Data are presented in arbitrary units after normalization of the raw quantitative PCR values with genes (ACTG1,
# EFF1A1, PNN and RHOT2) that presents highly similar values in all samples. Gene expression of the 16 genes are presented on FIGS. 4 and 5.

The 16-Gene Signature as a Strong Independent Prognostic Factor

In a First Set of 61 Patients

The clinical impact of HB molecular classification was addressed in a first set of 61 patients (FIGS. 7 and 8), comprising 37 (61%) C1 and 24 (39%) C2 cases. Besides strong association with predominant immature histotypes, HBs of the C2 subclass were tightly associated with features of advanced tumor stage, such as vascular invasion and extrahepatic metastasis (FIG. 6A). Accordingly, overall survival of these patients was markedly impaired. Kaplan-Meier estimates of overall survival probability at 2-years were 50% for patients with C2 tumors and 90% for patients with C1 tumors (p=0.0001, log rank test), and similar trends were seen for disease-free survival probabilities (data not shown). Next, we examined whether pre-operative chemotherapy treatment given to 48 patients could affect tumor classification. These cases were evenly distributed among HB subclasses, with no significant association with molecular classification. Of note, available pretreatment biopsies were assigned to the same subclass as matched resected tumors in 3 out of 4 cases (see FIG. 3; HB112 and HB112b have been both classified as C1 grade, and HB114 and HB114b have been both classified as C2 grade). We examined the performance of the 16-gene signature on the 48 tumors resected after chemotherapy, and found significant difference in outcome between patients with C1 and C2 type HBs (p=0.0021, log rank test) (FIG. 6B). Remarkably, Kaplan-Meier analysis confirmed C2 subclass as a poor prognostic group in 17 cases for which pre-treatment biopsies or primary surgery specimens were available (p=0.0318, log rank test) (FIG. 6C).

We further assessed the prognostic validity of the 16-gene signature for all patients in multivariate analysis, using a Cox proportional hazards model with pathological and clinical variables associated to patients' survival. This analysis identified the signature as an independent prognostic factor, with better performance than tumor stage defined by PRETEXT stage, vascular invasion and extrahepatic metastases (FIG. 6D). Thus, this signature demonstrated strong prognostic relevance when compared to current clinical criteria.

In a Second Set of 86 Patients

The clinical impact of HB molecular classification was addressed in a second set of patients (comprising the sample of the first set), comprising 53 (61%) C1 and 33 (39%) C2 cases. Besides strong association with predominant immature histotypes, HBs of the C2 subclass were tightly associated with features of advanced tumor stage, such as vascular invasion and extrahepatic metastasis (FIG. 9A). Accordingly, overall survival of these patients was markedly impaired. Kaplan-Meier estimates of overall survival probability at 2-years were 60% for patients with C2 tumors and 94% for patients with C1 tumors (p=0.00001, log rank test), and similar trends were seen for disease-free survival probabilities (Table 9).

Table 9 N. of patients 61 C1+25 C2 = 86 P value Survival (all patients) Alive/Dead C1 50/3 <0.00001 C2 20/13 DFS (all Datients) DFS/others C1 48/5 <0.00001 C2 18/15 Survival (non-treated Patients) Alive/Dead C1 12/0 0.0164 C2 11/6 DES (non-treated patients) DES/others C1 12/0 0.0213 C2 12/6
Survival analysis (Kaplan Mejer, log rank test); DES: disease-free survival; Others: dead or alive with recurrent disease.

Next, we examined whether pre-operative chemotherapy treatment given to 73 patients could affect tumor classification. These cases were evenly distributed among HB subclasses, with no significant association with molecular classification. We examined the performance of the 16-gene signature on the 73 tumors resected after chemotherapy, and found significant difference in outcome between patients with C1 and C2 type HBs (p=0.0002, log rank test) (FIG. 9B). Remarkably, Kaplan-Meier analysis confirmed C2 subclass as a poor prognostic group in 29 cases for which pre-treatment biopsies or primary surgery specimens were available (p=0.0164, log rank test) (FIG. 9C).

We further assessed the prognostic validity of the 16-gene signature for all patients in multivariate analysis, using a Cox proportional hazards model with pathological and clinical variables associated to patients' survival. This analysis identified the signature as an independent prognostic factor, with better performance than tumor stage defined by PRETEXT stage, vascular invasion and extrahepatic metastases (FIG. 9D).

Finally, various clinical elements of 103 HB samples from 86 patients were compared with respect to their classification as C1 or C2 grade using the 16-gene signature (Table 10).

TABLE 10 Clinical correlations. N. of patients 61 + 25 = 86 p-value (chi-square) Gender ns Chemotherapy treatment Yes/No C1 47/6 ns C2 26/7 Chemotherapy protocol STD/High C1 30/13 0.007 C2  9/16 TUMOR STAGE Early/Advanced C1 32/20 0.005 C2 10/23 Metastasis No/Yes C1 43/10 0.004 C2 17/16 Vascular Invasion No/Yes C1 36/15 0.005 C2 13/20 Advanced Pretext stage (IV) No/Yes C1 42/9 ns C2 24/7 Multifocality No/Yes C1 36/17 ns C2 18/14 Histology Ep/Mixed C1 31/21 ns C2 20/13 Main EDith ComD Fetal/NonFetal C1 48/4 <0.0001 C2  6/22
STD: standard risks (cisplatine) - High:high risk (cisplatine/doxorubicine, intensified treatment); Tumor stage (defined as Vasc. Inv and/or metastasis and/or PRETEXT stage IV); metastasis: extrahepatic metastasis (mainly lung); vascular invasion is determined by imagery; Pretext IV (involved an intrahepatic extent of the tumor to all hepatic sections);
# multifocality (more than 2 tumor nodules); Ep: pure epithelial form - Mixed: mesenchymatous and epithelial mixed form; Fetal: well differentiated; non fetal: embryonic, atypic, SCUD and/or macrotrabecular cells.

The above results carried out on a first set of 61 patients, and on a second completed set of 86 patients, demonstrate that the 16-gene signature, identified in the present application, is a strong prognostic relevance when compared to current clinical criteria.

Discussion

The present application demonstrates that, using integrated molecular and genetic studies, hepatoblastoma encompass two major molecular subclasses of tumors that evoke early and late phases of prenatal liver development. Aberrant activation of the canonical Wnt pathway represented a seminal event in both tumor types, with cumulated mutation rates of β-catenin, APC and AXIN over 80%. However, depending on tumor differentiation stage, Wnt signaling activated distinct transcriptional programs involved in tumor growth and invasiveness or in liver metabolism. Further comparisons of immature, embryonal-type HBs with the bulk of more differentiated, fetal-type tumors revealed a tight correlation between stage of hepatic maturation arrest and clinical behavior, notably vascular invasion and metastatic spread, and patients' survival.

Molecular Hb Subclasses are Determined by Liver Differentiation Stages

In this study, expression-based classification of HB was achieved through a highly reliable statistical method combining different unsupervised hierarchical clustering approaches. This method led to the selection of two robust tumor subgroups, and this robustness was confirmed using a new, independent set of samples and 16 relevant genes discriminating these tumor subgroups. These results demonstrated that the most significant differences between HB subclasses can be ascribed to distinct hepatic differentiation stages, as defined by comparison with expression profiles of mouse livers at early (E11.5-E12.5) and late (E14.5-E18.5) embryonic stages. These studies also provide biological relevance to early histologic classification that distinguished fetal and embryonal cells as major HB components (Weinberg and Finegold, 1983). The C1 subclass recapitulates liver features at the latest stage of intrauterine life, both by expression profile and by mostly fetal morphologic patterns, while in the C2 subclass, transcriptional program and predominant embryonal histotype resemble earlier stages of liver development. Thus, despite frequent morphological heterogeneity in HB, these expression-based subclasses closely matched the histologic types found to be prevailing after microscopic examination of the entire tumor mass.

These results, showing that childhood liver tumors recapitulate programs of their developing counterpart, are in line with recent studies using cross-species comparisons. It has been demonstrated that clinically distinct medulloblastoma subtypes can be identified by their similarity with precise stages of murine cerebellar development (Kho et al., 2004). Evidence for conserved mechanisms between development and tumorigenesis was also obtained in Wilms' tumor, the embryonic kidney malignancy, which shares expression of sternness and imprinted genes with murine metanephric blastema (Dekel et al., 2006). It was noticed that HBs, like Wilms' tumors, exhibit robust overexpression of a number of paternally expressed genes like DLK1, IGF2, PEG3, and PEG10 that are involved in growth induction processes and downregulated with differentiation during development.

Previous studies using stem cell markers and markers of hepatocytic and biliary lineages have described differential patterns among HB components that reflect sequential stages of liver development (Schnater et al., 2003). The present data extent these observations, and indicate that immature C2-type tumor cells evoke hepatic cancer progenitor cells, with distinctive overexpression of highly relevant markers such as cytokeratin 19 and Ep-CAM (Roskams, 2006). Recently, embryonic stem/progenitor cells have been isolated from human fetal livers, either by enrichment of blast-like cells in primary hepatoblast cultures or by immunoselection of Ep-CAM-positive epithelial cells (Dan et al., 2006; Schmeizer et al., 2007). These cell lines have self-renewal capacity and can differentiate into mature hepatocytes and cholangiocytes, and one of them also gives rise to various mesenchymal lineages (Dan et al., 2006). Whether HBs arise from transformation of these cell types is presently unknown. As malignant mesenchymal derivatives are frequently admixed with epithelial tissues in HB, it is tempting to speculate that this tumor occurs from a multipotent progenitor harboring characteristics of mesenchymal-epithelial transitional cells. Moreover, since no significant differences in gene expression profiles was noted here between pure epithelial and mixed epithelial-mesenchymal HBs, tumor cells likely kept intrinsic capacities to undergo epithelial-mesenchymal transition.

A salient feature of immature HBs is the characteristic interplay of sternness and proliferation found in aggressive tumors (Glinsky et al., 2005). The C2-type expression profile was significantly enriched in hESC markers, including the mitotic cell cycle and spindle assembly checkpoint regulators cyclin B1, BUB1, BUB1B, and Aurora kinases. These mitotic kinases are centrosomal proteins that ensure proper spindle assembly and faithful chromosome segregation in mitosis. Overexpression of these kinases or other components of the spindle checkpoint induces centrosome amplification and defects in chromosome segregation leading to chromosome number instability and aneuploidy (Marumoto et al., 2005; Zhou et al., 1998). Non-disjunctional events are involved in developmental syndromes (Hassold and Hunt, 2001), and might be responsible for increased rate of chromosomal imbalances evidenced here in C2-type HBs.

Context-Dependent Transcriptional Programs Driven by Wnt Signalling

Mutational activation of β-catenin is a hallmark of HB, and accordingly, we found intracellular accumulation and nuclear localization of the protein in virtually all tumors, albeit with variable frequencies and intensities. Both immature and differentiated tumors overexpressed AXIN2 and DKK1, reflecting an attempt to activate a negative feedback loop aimed at limiting the Wnt signal. However, the two HB subtypes showed significant differences in β-catenin immunoexpression, illustrated by concomitant nuclear accumulation and decreased membranous localization of the protein in poorly differentiated, highly proliferative HBs. Heterogeneous distribution of nuclear β-catenin within colorectal tumors has been linked to different levels of Wnt signaling activity, resulting from differential combinations of autocrine and paracrine factors (Fodde and Brabletz, 2007). Similarly, nuclear β-catenin might be related to the absence of membranous E-cadherin in immature HBs, as we reported previously (Wei et al., 2000), and to cross-talks with growth-stimulating pathways in less differentiated cells. In this context, increased dosage of Wnt signaling might induce migratory and invasive phenotype.

Major differences between the two HB subtypes were found here in expression levels of Wnt targets involved in liver functions. Recent studies have demonstrated that Wnt/β-catenin signaling governs liver metabolic zonation by controlling positively the perivenous gene expression program and negatively the periportal program (Benhamouche et al., 2006). In our study, overexpression of hepatic perivenous markers such as GLUL was prominent in differentiated HBs, while genes encoding periportal functions like GLS2 were downregulated. This profile is highly similar to those of human and murine HCCs expressing mutant β-catenin (Boyault et al., 2007; Stahl et al., 2005), and corresponds to an hepatic signature of Wnt target genes. Accordingly, the zonation-related profile was lessened in poorly differentiated HBs, and mutant β-catenin was found to activate a different, muscle-related expression program in the pediatric Wilms' tumor (Zirn et al., 2006).

Clinical Implications

The clinical behavior of many human solid tumors has been related to their differentiation status and proliferative rate. We show that HB does not depart from this rule, with strong correlation of molecular subclasses linked to hepatic differentiation with clinical tumor stage and patient's outcome. This correlation was mainly determined by differences in invasive and metastatic phenotypes between the two subclasses, but not by differences in tumor localization or tumor extension across liver sections, which defines the preoperative staging (PRETEXT) utilized to evaluate tumor resectability (Perilongo et al., 2000). Major differences in expression profiles of the two molecular HB subtypes led us to elucidate a 16-gene signature that proved highly efficient in stratification of HBs as well as normal livers according to hepatic developmental stage. Most importantly, this classifier also discriminated aggressive tumors, exhibited powerful survival predictor capacities in pre-treatment biopsies and surgical specimens, and demonstrated strong prognostic relevance when confronted to current clinical criteria in multivariate analysis. Although immature HBs have been associated to worse clinical outcome as opposed to differentiated HBs (Weinberg and Finegold, 1983), frequent cellular heterogeneity has hampered the use of histopathologic criteria for defining risk groups, excepted for a minority of cases showing ‘pure fetal’ or SCUD types. The expression signature afforded here enables direct appraisal of the global degree of tumor cell maturation, allowing to bypass these difficulties. Thus, it can improve the outcome prediction and clinical management of hepatoblastoma, by identifying cases with increased risk of developing metastasis, or conversely, by avoiding unnecessary over-treatment.

In conclusion, the present application identifies a 16-gene signature that distinguishes two HB subclasses and that is able to discriminate invasive and metastatic hepatoblastomas, and predicts prognosis with high accuracy. The identification of this expression signature with dual capacities may be used in recognizing liver developmental stage and in predicting disease outcome. This signature can be applied to improve clinical management of pediatric liver cancer and develop novel therapeutic strategies, and is therefore relevant for therapeutic targeting of tumor progenitor populations in liver cancer.

Analysis of 64 Hepatocellular Carcinoma (HCC) from 64 Patients

Real time RT-PCR (Taqman methodology) was performed on 67 HCC samples, as disclosed for HB samples above. The clinical characteristics of the 67 patients diagnosed with HCC as well as the features of the HCC samples are disclosed in Tables 11 and 12 below.

Amplification was carried out with primers of the 16-gene signature disclosed in Table 6. Data were normalized to the expression of the ROTH2 gene (primers disclosed in Table 7) and analyzed by the ΔCt method. Quantitative PCR data are disclosed in Table 13.

TABLE 11 features of the HCC samples obtained from 67 patients (pages 60 to 62) Tumor follow-up tumor grade tumor differentiation tumor vascular invasion recurrence or Id length (years) (Edmonson) according to OMS size macro micro metastasis HC1 0.07 3 moderately differentiated 120 NA absent no recurrence HC10 0.95 4 moderately/poorly differentiated 75 absent absent no recurrence HC11 11.10 NA NA 15 absent absent no recurrence HC12 0.05 NA Well differentiated 60 NA NA no recurrence HC14 1.00 NA moderately/poorly differentiated 80 NA NA no recurrence HC15 1.22 3 moderately differentiated 60 present present no recurrence HC17 10.96 2 Well differentiated 100 absent absent no recurrence HC18 0.39 3 moderately differentiated 140 present present NA HC20 15.40 NA Well differentiated 40 NA NA no recurrence HC21 0.70 NA NA 100 NA NA NA HC22 11.50 NA Well differentiated 45 absent absent no recurrence HC23 11.93 2 Well differentiated 50 absent absent no recurrence HC25 15.87 2 Well differentiated 140 absent absent NA HC27 0.10 NA Well differentiated 15 absent absent no recurrence HC28 0.10 NA moderately differentiated 120 NA present no recurrence HC3 3.33 2 Well differentiated 60 absent absent recurrence HC30 11.78 3 moderately differentiated 16 NA NA no recurrence HC32 0.66 2 Well differentiated 60 absent NA no recurrence HC34 14.72 2 Well differentiated 140 absent absent recurrence HC37 0.20 NA moderately differentiated 35 present present non HC38 1.12 NA NA 50 absent NA recurrence HC4 11.48 2 Well differentiated 100 absent absent no recurrence HC41 7.44 2 Well differentiated 30 NA absent recurrence HC42 10.58 3 moderately differentiated 130 possible; present no recurrence non certain HC43 10.20 NA moderately differentiated 15 NA NA no recurrence HC52 0.25 3 moderately differentiated 110 absent absent no recurrence HC58 8.30 2 moderately differentiated 100 absent absent no recurrence HC6 1.25 2 Well differentiated 90 absent present recurrence HC64 5.25 3 moderately differentiated 40 absent absent recurrence HC66 8.93 2-3 Well to moderately differentiated 75 absent absent no recurrence HC7 1.50 2-3 Well differentiated 100 present present recurrence HC8 8.48 3 moderately differentiated 30 absent absent no recurrence HC9 0.02 3-4 moderately/poorly differentiated 100 present present no recurrence HC101 1.00 2-3 Well to moderately differentiated 35 present present no recurrence HC102 0.10 NA Poorly differentiated 200 present present no recurrence HC103 1.82 2-3 Well to moderately differentiated 55 absent present recurrence HC104 0.17 2-3 Well to moderately differentiated 160 Possible; present no recurrence non certain HC105 0.56 3 moderately differentiated 40 present present recurrence HC106 1.70 3 moderately differentiated 80 present present no recurrence HC107 1.75 2 Well differentiated 60 absent absent no recurrence HC108 1.62 3 moderately differentiated 26 absent present no recurrence HC109 1.00 1-2 Well to very well differentiated 30 absent absent no recurrence HC110 1.00 3 moderately differentiated 30 present present no recurrence HC111 0.60 3 moderately differentiated 40 present present no recurrence HC1112 1.48 2-3 Well to moderately differentiated 18 absent absent no recurrence HC113 1.00 2-3 Well to moderately differentiated 50 present present no recurrence HC114 0.44 2 Well differentiated 36 absent absent no recurrence HC119 0.75 1 Well differentiated 90 absent absent no recurrence HC120 0.69 3 moderately differentiated 140 absent absent no recurrence HC121 1.00 2-3 Well to moderately differentiated 28 absent absent no recurrence HC122 0.93 1 Very well differentiated 40 absent absent no recurrence HC123 0.90 3 moderately differentiated 26 absent present no recurrence HC124 0.82 2-3 Well to moderately differentiated 20 absent present no recurrence HC125 0.60 3 moderately differentiated 150 Possible; present no recurrence non certain HC126 0.75 2 Well differentiated 20 present present recurrence HC127 0.40 3 moderately differentiated 43 probable probable no recurrence HC128 0.52 3 moderately differentiated 62 absent absent no recurrence HC129 0.30 3 moderately differentiated 25 absent present no recurrence HC131 0.42 1-2 Well differentiated 130 present present recurrence HC132 0.25 2-3 Well to moderately differentiated 115 present present recurrence HC133 0.44 2 Well to moderately differentiated 110 absent present no recurrence HC134 0.10 3 moderately differentiated 30 absent present no recurrence HC135 0.14 3 moderately differentiated 38 absent Possible; no recurrence non certain HC136 0.26 2-3 Well to moderately differentiated 120 absent present no recurrence
N.A: non available;

macro: macrovacular invasion;

micro: microvacular invasion

TABLE 12 features of the HCC samples obtained from 67 patients, and features of patients (pages 63 and 64) Chronic Other Tumor Score METAVIR viral Viral etiology etiolo- ID Activity Fibrosis hepatitis HBV HCV alcohol gies HC1 NA 4 no no no yes HC10 NA 4 yes yes no no HC11 NA NA yes yes yes no HC12 NA NA yes yes no no HC14 NA NA yes no yes yes HC15 3 3 no no no yes HC17 NA 3 yes yes no no HC18 2 4 no no no yes HC20 NA NA no no no yes HC21 NA NA no no no yes HC22 NA NA no no no yes HC23 NA 0 no no no no HC25 0 0 no no no no HC27 NA NA yes no yes no HC28 0 0 no no no no HC3 NA 4 yes no yes no HC30 NA 4 no no no yes HC32 NA 4 yes no yes no HC34 NA 0 no no no no HC37 NA NA no no no yes HC38 NA 4 yes no yes no HC4 NA 1 no no no no HC41 NA 4 yes no yes no HC42 2 1 yes yes no no HC43 NA NA yes no yes no HC52 NA 4 yes yes no no HC58 2 3 yes no yes no HC6 NA 1 no no no yes Hemochro HC64 2 2 yes no yes no HC66 NA 4 yes yes no yes HC7 2 3 no no no yes HC8 NA 4 yes no yes no HC9 1 3 no no no yes HC101 2 4 yes yes yes yes HC102 1 1 yes yes yes no HC103 3 4 yes yes no no HC104 0 1 no no no no HC105 2 4 yes no yes no HC106 1 4 yes yes no no HC107 0 0-1 no no no yes HC108 1 1 yes no yes no HC109 2 4 no no no yes NASH HC110 1 4 yes no yes yes HC111 1 4 no no no yes HC112 2 2 no no no no NASH HC113 1 4 yes no yes no HC114 2 3 no no no yes HC119 2 1 no no no no NASH HC120 2 3 yes yes no no HC121 2 4 yes no yes no HC122 0 1 no no no no HC123 2 4 yes no yes yes HC124 1 4 yes yes no no HC125 2 4 no no no yes NASH HC126 1 4 yes yes no no HC127 2 4 yes no yes no HC128 1 1 no no no no NASH HC129 2 4 no no no yes HC131 0 1 no no no no HC132 1 1 yes yes no no HC133 2 2 no no no yes HC134 2 3 yes no yes no HC135 1 2 yes yes no no HC136 0 1 no no no no
N.A: non available; HBV: hepatitis B virus; HCV; hepatitis C virus; hemochro: hemochromatosis; NASH non alcoholic steatohepatitis.

TABLE 13 Quantitative PCR data of the 16-gene signature normalized to the expression of the ROTH2 gene (pages 65 to 68) HC1 HC3 HC4 HC6 HC7 HC8 HC9 HC10 HC11 AFP -2.212911 -3.865709 -7.6758115 -7.9469815 5.311541 2.0890815 -70483095 2.3869635 0.6488335 ALDH2 6.2372335 6.230074 2.186358 5.4231035 4.0446765 3.9297005 3.0017225 0.95212 5.958108 AP0C4 0.614689 0.95786 -1.608247 0.9614255 -3.550537 -0.6776965 -9.6721075 NA 1.076151 APCS 7.0721355 7.52919 5.845683 7.3704745 5.1967915 6.567126 -0.017488 -1.0272875 7.7638255 AQP9 6.047695 6.7334475 3.759528 7.006052 6.747103 3.1082155 3.7536735 1.3400495 6.122144 BUB1 -3.841505 -0.147459 -4.221132 -0.5252045 -0.299039 -1.214781 2.980029 -1.864677 -2.362454 C1S 8.163492 8.7963405 5.8997645 8.162856 4.062593 7.2991535 4.830331 2.639902 8.319293 CYP2E1 10.3093235 10.428074 7.1147515 10.1334265 11.024027 7.7910075 0.5825245 3.604805 9.575619 DLG7 -5.30317 -2.057513 -4.4226465 -1.6282005 -1.169221 -2.80866 1.3733475 NA -2.8432205 DUSP9 -11.616567 -8.8462855 -9.4268185 -10.22051 -6.6521625 -9.6946695 -9.5262655 NA NA E2F5 0.05328 -1.909804 -1.7432195 0.024339 -0.2833465 -0.0193165 0.711082 -1.344368 -0.736822 GHR 2.655512 2.069524 -2.0012965 1.887805 -1.7428205 2.342442 -2.3242195 -0.4900285 4.757848 HPD 9.449416 8.549803 9.415253 8.5958965 6.183977 5.329776 -0.011478 2.932809 9.029214 IGSF1 -6.46034 -7.249974 NA -7.1580385 -3.192514 -2.806768 -4.026769 NA -7.6390015 NLE1 -1.159417 -1.5801355 -3.1459935 0.6940375 -0.3919565 -1.579419 -0.80375 NA -1.9328755 RPL10A 6.6225235 6.0562915 4.4121905 6.8637555 7.1381125 6.2574845 6.3016635 9.1966395 7.379063 HC12 HC15 HC17 HC18 HC20 HC21 HC22 HC23 HC25 AFP -6.538312 6.14089 7.1950405 -6.856588 -0.65281 -4.3070475 -4.418018 -5.538438 -3.90298 ALDH2 4.6271565 4.5178635 2.6522585 1.840894 6.287083 2.175112 5.331214 5.853486 6.162477 AP0C4 -1.221393 -5.156026 -2.395651 -3.84764 3.2094885 -6.2591235 0.5455545 0.5708905 1.834891 APCS 6.942673 3.380102 4.5167035 4.916924 8.2117635 5.9159775 6.6835035 6.9009145 8.798759 AQP9 4.1878425 2.373344 2.8711295 3.6093495 7.354605 1.1452535 5.7992305 6.651868 8.758959 BUB1 -3.293346 0.8830545 1.0884485 -0.063545 -1.4635025 0.0802935 -2.173361 -2.5475915 -2.5679685 C1S 6.850023 7.1343975 6.035123 4.263272 8.471663 5.7190985 7.2514145 8.2212235 8.5606875 CYP2E1 7.284587 4.9390935 6.037085 5.811062 10.2536915 1.2878015 8.0876755 9.047509 10.814935 DLG7 -4.7199665 -0.1414205 0.666284 -1.512286 -2.1165725 -0.322455 -3.3904095 -3.848364 -3.34202 DUSP9 NA -4.4342765 -3.163581 -8.7756845 -9.6208445 -7.8162765 -10.827291 NA -7.1111525 E2F5 -2.4002515 1.399564 1.206766 -2.426129 -1.1944835 -0.0686475 -0.7133385 -1.4330655 0.049846 GHR 2.2402875 0.2426 -2.353691 -2.9035 4.5756335 0.71981 2.416651 3.7226655 1.9012935 HPD 9.656029 4.473096 0.6808655 5.7101575 10.6864405 4.0108195 9.8859985 9.583194 9.1845675 IGSF1 -7.466951 0.0722075 -6.0490105 -2.4248235 NA -2.954514 -5.6986975 -7.200325 NA NLE1 -1.64183 -0.321593 -0.386649 -1.3815525 -1.118745 -1.618369 -1.9449755 -1.823275 -1.770127 RPL10A 5.178571 6.8777395 7.068098 5.9464565 7.542193 6.309556 7.194012 5.9526365 7.4507165 HC26 HC27 HC28 HC30 HC32 HC34 HC37 HC38 HC41 AFP -5.69175 -0.626755 NA 6.4370325 0.0037145 -6.6945705 -1.3519745 4.053435 -2.7156435 ALDH2 5.0135775 5.6309605 1.913778 3.8476295 6.802666 5.11617 5.808058 4.596143 6.3503265 AP0C4 0.2581675 1.53158 -6.0251725 0.2797975 2.574347 0.5860455 -0.0768065 -0.129322 2.281983 APCS 7.2072275 7.2809855 1.0475505 7.1142435 7.500133 7.134934 6.755895 5.045701 5.612517 AQP9 3.8645965 5.4736555 0.9613895 5.0250435 7.530391 6.9427395 6.3416265 6.0302545 7.8444565 BUB1 0.545363 -0.8889165 -5.7426525 -0.190936 -5.1317805 -1.2674215 -2.4955985 0.321483 -0.587016 C1S 7.2351705 8.172076 4.910584 7.5279395 7.854502 7.719763 6.921051 6.101331 6.88808 CYP2E1 0.671071 8.6350095 3.6858305 7.5682115 9.4408715 8.545814 10.1686795 8.1123675 9.5090495 DLG7 -0.9710395 -2.3158215 NA -0.189092 -5.7080765 -2.339621 -2.6534895 -1.4386515 -1.840185 DUSP9 -8.5287915 -10.241011 NA -9.0027 -9.73163 -9.9728495 NA -5.2298755 -8.727439 E2F5 -1.1845665 -0.4045835 -4.334386 1.0623035 -0.054818 -1.4281575 -1.2212655 -0.037887 0.466649 GHR 1.964045 2.623084 -1.9788575 2.635437 2.0027475 1.563203 2.9415775 0.2025015 1.428749 HPD 7.6403735 9.597772 3.3142495 7.537 9.0015185 8.3685675 10.367265 7.547286 8.0015745 IGSF1 -5.4960635 -5.588995 NA -2.651022 NA -10.112616 -7.5570255 -0.680358 -7.243446 NLE1 -1.851733 -1.851285 -2.4559905 -1.2674865 -1.208576 -1.934745 -1.9881245 -2.1250395 -0.15624 RPL10A 5.9670715 7.6623025 5.521873 7.5046195 8.8437815 6.594006 6.901637 5.1574215 7.7043325 HC42 HC43 KC44 HC52 HC58 HC60 HC64 HC66 HC101 AFP -5.216493 -1.7983435 -0.564605 10.3337105 1.891958 7.624821 5.0266755 3.156328 -6.873135 ALDH2 4.4086495 5.457548 7.1344115 2.1920375 2.1172735 3.6860195 4.992107 3.8408415 4.339036 AP0C4 -0.627239 -0.7055185 0.499817 -8.124407 -11.8524 -0.545509 0.7860345 -0.6773785 -0.5787185 APCS 4.1054755 7.607914 7.567581 5.9818015 -4.1106695 8.100997 7.4148835 8.2106815 6.288568 AQP9 6.063786 4.7175855 6.058158 -0.4848805 -2.817265 6.8503395 7.0526325 6.2767975 4.6233735 BUB1 -2.224818 -2.8634735 -3.5668895 -1.2986035 1.9395175 -0.576028 -1.367463 -1.1272665 0.081457 C1S 6.3060565 7.9862115 8.547705 5.6337865 3.691331 8.167253 7.1364365 8.026875 7.321092 CYP2E1 9.1411555 8.760714 9.1133175 1.7693015 -4.3317445 9.1875325 9.682147 8.601088 5.806032 DLG7 -3.2531575 -4.2390495 -4.814388 -2.599359 0.1957495 -2.2644225 -2.386875 -2.7680135 -1.3084655 DUSP9 NA -10.525647 NA -3.8059605 -3.656912 -6.618755 -7.3184655 -11.5673955 -8.828389 E2F5 -0.3673235 -0.894345 -1.894272 0.4419525 0.804087 -0.432422 -0.2876185 -0.968982 -1.871516 GHR -1.2545195 3.2916395 4.5598275 -1.843696 -3.7242975 -1.4079225 0.349645 -1.2501855 0.1466275 HPD 8.2669835 8.997825 9.158005 2.481945 1.8257985 8.4643875 8.6027575 8.5231325 5.7252795 IGSF1 -2.899766 -5.5544715 -5.769786 2.254168 1.3471695 -0.7884805 -3.3382005 -9.185554 -4.1394545 NLE1 -0.9401045 -1.8422595 -2.0303285 -1.9474305 -1.209522 -1.9133155 -1.817699 -1.962008 -1.4546305 RPL10A 5.577659 5.480403 5.8488475 5.6154705 6.0601515 5.7041285 6.4617635 5.415169 6.144011 HC102 KC103 HC104 HC105 HC106 HC107 HC108 HC109 HC110 AFP -4.119697 1.6193685 5.5094265 2.3444245 -3.42054 -4.136209 -4.500336 -4.833024 -3.5240185 ALDH2 2.476355 3.889904 4.936239 4.239726 6.1642895 6.7443095 3.6076385 5.8617665 3.6707715 AP0C4 -5.453696 -0.54698 -0.5059805 -3.577778 -0.7836775 4.4534435 -2.478085 0.729565 -0256479 APCS -2.3952165 6.014572 5.624234 7.703333 7.8462545 9.2080655 7.275462 6.222909 5.043319 AQP9 0.0196725 7.151639 0.501258 4.2748785 5.85931 8.8878655 4.4353395 6.4504115 4.5999895 BUB1 -0.5553155 -2.086008 -1.311194 0.945674 -4.8909655 -1.7415115 -0.3807995 -2.2918285 -1.449943 C1S 5.939374 5.965432 6.716137 7.774455 8.060072 9.2061165 7.1031155 7.406001 6.9163195 CYP2E1 -2.8566735 8.266311 9.0888685 5.698899 9.9949555 9.3234825 3.889942 8.7101925 7.1 45766 DLG7 -2.1385165 -2.957914 -1.821739 -0.814912 -6.2678815 -1.357756 -2.2445545 -3.222524 -2.333076 DUSP9 -8.6628475 -12.521336 -5.396553 -5.4214725 -11.174152 -6.6136855 -8.0946735 -10.4709205 -11.616244 E2F5 0.830934 -1.8003215 -2.305498 2.0730715 -2.208171 2.78876 0.0923905 -1.9924345 -2.512512 GHR 0.947389 0.636723 1.6860905 0.682142 5.342392 2.935929 1.6363755 2.9233285 1.0803015 HPD 0.568809 6.717282 8.46781 2.288109 9.4440475 10.460972 2.9674235 7.8859205 8.1908235 IGSF1 -2.708733 -9.802921 0.1438735 -1.422332 -7.401009 NA -7.967992 -10.0122565 -8.1469415 NLE1 -1.1534675 -2.594702 -1.610158 -0.471391 -1.968983 -0.000835 -0.932052 -2.6102395 -2.3529485 RPL10A 5.283399 4.423835 6.21159 6.315756 5.769397 8.6686655 5.818028 5.541229 5.245476 HC111 HC112 HC113 HC114 HC119 HC120 HC121 HC122 HC123 AFP -1.883473 -2.8803905 1.208649 -5.4433695 1.0580855 -4.0065425 -4.254961 -2.3763095 0.821555 ALDH2 3.8304065 4.8726745 4.407016 4.7113965 6.159706 4.257398 4.556431 6.2844515 4.220769 AP0C4 -1.130067 -0.7777655 -2.366969 -0.833543 1.894453 -3.5241745 -2.167313 1.279577 -0.68167 APCS 5.976754 6.764675 5.197177 6.723142 9.375177 5.6838965 6.2688205 6.9942545 5.778659 AQP9 4.1657805 5.2735435 2.681192 4.445291 7.6266135 6.8239115 4.38702 6.8198535 6.410177 BUB1 0.621548 0.3135015 -3.4825665 -1.7431855 -0.797564 -0.0740105 -2.4486685 -6.0183915 -1.190323 C1S 6.278164 7.455794 6.338901 7.866014 9.1461175 8.5708615 8.118416 7.7653135 5.383781 CYP2E1 4.46942 2.5741475 6.443846 7.3429245 7.095824 7.6044515 7.765037 9.450349 8.528543 DLG7 -0.769283 -0.9196845 -4.5602875 -3.1500875 -1.712686 -1.9563135 -2.852561 -7.228946 -2.929576 DUSP9 -9.137462 -10.105965 -7.8299455 -11.804112 -9.106547 -5.8119685 -9.706684 -9.9054825 -11.584458 E2F5 1.045678 0.0373705 -2.82243 -0.0450475 -0.0248045 1.229768 -0.910943 -3.5033365 -0.646839 GHR 1.1576425 2.5391085 2.16232 2.5053965 3.7649595 3.196589 2.2774645 2.400201 -1.810364 HPD 7.245347 7.714358 6.685692 6.835254 9.220498 8.5127155 7.480725 8.7301975 4.7774665 IGSF1 -1.86965 -3.4428695 -2.045068 -5.1813245 -5.39017 -9.404196 -5.980435 -8.6480295 -5.1400615 NLE1 -1.012752 -1.119237 -2.156348 -1.3170345 -0.400823 -1.1096815 -1.758163 -2.2430545 -1.5951645 RPL10A 5.568205 6.1905075 5.8884625 5.795905 7.954231 6.4517175 6.4042545 5.199782 4.7323885 HC124 HC125 HC126 HC127 HC128 HC129 HC131 HC132 HC133 AFP 3.9525335 -4.806564 -5.899437 -0.0390765 5.8636305 -3.430757 -1.491189 5.4265205 -5.1621395 ALDH2 4.027289 4.5451465 5.02839 2.41699 5.085525 4.6298475 5.425994 3.105643 4.2462915 AP0C4 -0.0499065 2.6326775 0.407895 0.8680995 -0.626498 -1.863955 2.4702 -6.9974515 0.63156 APCS 5.391271 6.5321595 5.2838365 4.846116 5.087517 4.8448705 8.6617295 -3.2748865 7.145861 AQP9 4.463488 8.370224 3.6163545 1.8613935 4.3184915 2.870839 7.4772145 3.9244375 6.05182 BUB1 -1.592563 1.1627945 -2.6943025 -2.048769 -1.3297375 -2.3688215 -0.727709 0.2895395 -4.9277675 C1S 5.151686 8.4244055 7.1365955 6.3641695 6.828468 7.302922 7.525072 4.390082 7.3188145 CYP2E1 9.520436 9.426232 5.226091 6.1813065 7.4344035 2.692798 8.98645 7.0455735 8.1908895 DLG7 -2.03781 0.3286545 -3.944339 -2.96212 -2.6299155 -3.6405185 -1.461713 -1.5572645 -5.5447335 DUSP9 -8.81055 -9.3740615 -8.7174575 -8.672372 -8.499355 -7.0627455 -8.415907 -3.3843145 -8.022457 E2F5 0.574165 -0.028878 -3.271927 -2.162802 -4.393094 -0.470421 0.154573 1.9018925 -2.6341525 GHR 2.2369305 0.697866 1.824385 0.129431 1.9716885 2.332961 4.009655 1.7710325 2.2298335 HPD 7.832169 5.7813 1.865621 3.4481965 5.7052855 5.502918 8.960383 2.3653865 6.1281315 IGSF1 -1.4450915 -10.2234745 -7.659377 -3.1503205 -2.72995 -5.692623 -7.5832005 -1.947055 NA NLE1 -0.1499775 -0.405397 -2.033278 -2.205965 -1.949352 -1.683808 -1.5313675 0.2035885 -1.4173895 RPL10A 6.691521 7.1196575 5.389272 4.3385115 6.6181545 4.8697295 6.775249 6.7796075 5.762015 AFP HC134 HC135 HC136 ALDH2 2.8738695 -0.909107 -0.4105125 AP0C4 4.061101 2.7442165 6.0408575 APCS -0.1134065 -0.7630605 0.7390785 AQP9 7.5103485 0.959726 7.150737 BUB1 5.550642 4.0595615 5.996196 C1S 1.7425995 -1.2018365 -4.288554 CYP2E1 8.4609335 4.667223 8.243333 DLG7 7.859701 4.30592 9.042865 DUSP9 0.8148735 -2.250305 -5.5267715 E2F5 -4.96739 -5.794605 -10.9307725 GHR 3.1030595 0.986165 -2.4040865 HPD 1.3138565 -0.6955465 4.013948 IGSF1 7.231144 6.7262275 8.223611 NLE1 -0.3848995 -4.394354 -7.4962365 RPL10A 0.794433 -0.9780515 -2.426321 AFP 7.7140665 6.689595 5.5069335
NA: non available

Data were then analyzed by unsupervised clustering (dCHIP software) using 2 methods: average and centroid. Tumors were clustered into 2 groups, C1 and C2. Most of the samples have been attributed the same classification using the 2 methods, except for 6 samples (9%) that have been attributed a different classification (Table 15).

Clinical Parameters Associated to the C1 and C2 Molecular Subclasses

The clinico-pathological parameters of patients and tumors were compared between the two groups C1 and C2, using student's t test and Kaplan-Meier estimates. Since some data are not available for 3 patients, the following statistical studies were performed on 64 tumors.

Survival Analysis

There is a strong correlation of the molecular classification into C1 and C2 with patient's survival by using both classifications (Log rank: Centroid p=0.020 and Average p=0.024) (FIG. 10). In this figure, censored cases indicate the end of the follow-up (the last visit) for individual cases. Probability of survival at two years is 78% for C1 subclass and 39% for C2 subclass (the follow-up may be less than 2 years for some patients).

Association of HCC Classification with Clinical Variables

Table 14 shows the correlation between some clinical variable and the classification of the tumors.

TABLE 14 Variable C1 C2 p-value Tumor grade >2 (Edmonson) 13/29 21/23 <0.0001 Moderately-poorly differentiated (OMS) 17/36 23/25 <0.0001 Macrovascular Invasion  6/30  9/21 0.074 Microvascular Invasion 13/32 15/22 0.043 Recurrence 7/36  5/25 ns
(ns: non-significant)

TABLE 15 Classification of samples by unsupervised clustering (dCHIP software): average and centroid methods. Tumor ID average centroid comparison HC1 C1 C1 Same HC10 C2 C2 Same HC11 C1 C1 Same HC12 C1 C1 Same HC14 C1 C1 Same HC15 C2 C2 Same HC17 C2 C2 Same HC18 C2 C2 Same HC20 C1 C1 Same HC21 C2 C2 Same HC22 C1 C1 Same HC23 C1 C1 Same HC25 C1 C1 Same HC26 C1 C2 Different HC27 C1 C1 Same HC28 C2 C2 Same HC3 C1 C1 Same HC30 C2 C2 Same HC32 C1 C1 Same HC34 C1 C1 Same HC37 C1 C1 Same HC38 C2 C2 Same HC4 C1 C1 Same HC41 C1 C1 Same HC42 C2 C1 Different HC43 C1 C1 Same HC44 C1 C1 Same HC52 C2 C2 Same HC58 C2 C2 Same HC6 C1 C1 Same HC60 C2 C2 Same HC64 C2 C2 Same HC66 C1 C1 Same HC7 C2 C2 Same HC8 C2 C2 Same HC9 C2 C2 Same HC101 C1 C2 Different HC102 C2 C2 Same HC103 C1 C1 Same HC104 C2 C2 Same HC105 C2 C2 Same HC106 C1 C1 Same HC107 C1 C1 Same HC108 C1 C1 Same HC109 C1 C1 Same HC110 C1 C1 Same HC111 C2 C2 Same HC112 C1 C2 Different HC113 C2 C2 Same HC114 C1 C1 Same HC119 C1 C1 Same HC120 C1 C1 Same HC121 C1 C1 Same HC122 C1 C1 Same HC123 C2 C1 Different HC124 C2 C2 Same HC125 C1 C1 Same HC126 C1 C1 Same HC127 C2 C2 Same HC128 C2 C2 Same HC129 C1 C2 Different HC131 C1 C1 Same HC132 C2 C2 Same HC133 C1 C1 Same HC134 C2 C2 Same HC135 C2 C2 Same HC136 C1 C1 Same

In a second analysis, the global set of 64 tumors was analyzed independently of the C1/C2 classification, for parameters associated to survival. Results are presented in Table 16.

TABLE 16 Variable Log rank Tumor grade >2 0.108 Mod-poor Duff. Degree 0.400 Macrovasc. mv. 0.004 Microvasc. mv. 0.026 recurrence ns Tumor size 2cm 0.397 Score METAVIR Activity ns Score METAVIR Fibrosis 0.038 <2 vs. ≧ 2 (variable 3) Chronic hepatitis 0.948 HBV 0.093 HCV 0.352 Alcohol 0.225
(ns: non-significant)

These results demonstrate that the methods and the signatures of the invention are able to determine the grade not only of HB tumors but also of HCC tumors. The inventors have shown that hierarchical clustering is an efficient method for classification of tumor grade especially for HB. For HCC, this method may be less sufficient (less robust) when the amplitude of variation of expression results of the genes is less important than for HB.

Classification of Hepatoblastomas and Hepatocellular Carcinomas Using the Method of Discretization of Continuous Values.

85 hepatoblastomas (HBs) and 114 hepatocellular carcinomas (HCCs) including to the samples used in the above examples have been analyzed by quantitative PCR using the 16-gene signature and have been classified by the method of discretization of continuous values in order to determine their tumor grade.

Description of the Methodology for Classification

The inventors have designed a methodology for classification based on the principle of discretization of continuous values which refers to the process of converting continuous variables to “discretized” or nominal sets of values.

The major advantage of the discretization method relies on the definition of a cut-off for codification of each qPCR value (either by the Taqman or by the SybrGreen method), which provides an intrinsic score to directly classify an individual sample. There is hence no requirement to compare a sample to a large series of samples. In contrast, in other classification methods, the assigned subclass (such as C1 or C2 disclosed herein) is relative to the values obtained in a large number of cases. Moreover, the use of the average discretized values allows to tolerate missing values when analyzing the qPCR results (i.e. missed amplification of one of the genes for technical reasons).

Using the qPCR data of the 16 genes normalized to the reference RHOT2 gene (−deltaCt values), a cut-off (or threshold) has been defined for each gene. The −deltaCt values are converted into discrete values “1” or “2” depending on an assigned cut-off. In order to privilege the identification of samples that display strong overexpression of proliferation-related genes and/or strong downregulation of differentiation-related genes, the cut-offs have been defined as follows:

for the 8 proliferation-related genes (AFP, BUB1, DLG7, DUSP9, E2F5, IGSF1, NLE1, RPL10A), all −DeltaCts with a value above the 67th percentile have been assigned discretized value “2”, otherwise the assigned value was “1”.

for the 8 differentiation-related genes (ALDH2, APCS, APOC4, AQP9, C1S, CYP2E1, GHR, HPD), all −deltaCts with a value below the 33rd percentile have been assigned discretized value “1”, otherwise the assigned value was “2”.

Classification of 85 Hepatoblastomas (HB)

RNA Preparation and Quantitative PCR

RNA was extracted by using either Trizol, RNeasy kit (QIAGEN) or miRvana kit (Ambion), then quantified and quality-checked by Agilent technology.

For quantitative PCR analysis, the Sybr Green approach was used as described in point E. above. For each cDNA preparation, 1 μg of RNA was diluted at the final concentration of 100 ng/μl, and reverse transcribed with the Superscript RT kit (Invitrogen) following the manufacturer's protocol. Random primers were added at the final concentration of 30 ng/μl and the final volume was 20 μl. The cDNA was diluted 1:25, and 5 μl were used for each qPCR reaction. We added 5 μl of 2×Sybr Green Master mix (Applied Biosystems) and 0.3 μl of each specific primer (disclosed in point H. above) (final concentration 300 nM). Each reaction was performed in triplicate. qPCR reactions were run on the Applied Biosystems 7900HT Fast Real-Time PCR System with a 384-well thermo-block, and the conditions were the following:

2 min at 50° C. to activate Uracil-N-glycosylase (UNG)-mediated erase of a specific reaction

10 min at 95° C. to activate the polymerase and inactivate the UNG

40 cycles:

15 sec at 95° C. denaturation step

1 min at 60° C. annealing and extension

Final dissociation step to verify amplicon specificity.

The normalized qPCR (deltaCt) values of the 85 HB samples are given in Table A.

Analysis of qPCR Data.

Assignment of a discretized value for the 8 proliferation-related genes (“AFP” “BUB1” “DLG7” “DUSP9” “E2F5” “IGSF1” “NLE” “RPL10A”) was based on the 67th quantile (i.e. percentile), given that around ⅓ of HB cases overexpress proliferation genes, which is correlated with tumor aggressiveness and poor outcome. Assignment of a discretized value for the 8 differentiation-related genes (“ALDH2” “APCS” “APOC4” “AQP9” “C1S” “CYP2E1” “GHR” “HPD”) was based on the 33rd quantile, given that around ⅓ of HB cases underexpress differentiation genes, which is correlated with tumor aggressiveness and poor outcome.

The cut-offs (or thresholds) selected for the −deltaCT value of each gene were determined after considering said chosen percentiles for each group of genes are as follows:

AFP: 3.96139596; ALDH2: 4.3590482; APCS: 4.4691582; APOC4: 2.03068712; AQP9: 3.38391456; BUB1: −1.41294708; C1S: 4.24839464; CYP2E1: 6.70659644; DLG7: −3.3912188; DUSP9: 2.07022648; E2F5: −0.72728656; GHR: −0.1505569200; HPD: 2.27655628; IGSF1: 0.1075015200; NLE: −0.02343571999; RPL10A: 6.19723876

For the sample, the relative expression value is determined for each gene of the set of profiled genes. Each value is compared to the cut-off for the corresponding gene and is then discretized as a result of its position with respect to said cut-off.

The next step consisted in assigning a discretized score to each sample as follows:

1—the average of the “discretized” values of the 8 proliferation-related genes was determined. The 8 proliferation-related genes are the following: AFP, BUB1, DLG7, DUSP9, E2F5, IGSF1, NLE, and RPL10A.

2—the average of the “discretized” values of the 8 differentiation-related genes was determined. The 8 differentiation-related genes are the following: ALDH2, APCS, APOC4, AQP9, C1S, CYP2E1, GHR, and HPD.

3—The score for each sample was determined as the ratio between the average of proliferation-related genes and the average of differentiation-related genes.

According to this calculation, a score of 2 is the maximal score for highly proliferating and poorly differentiated tumors, whereas well differentiated and slowly proliferating tumors will have a minimal score of 0.5.

Based on the scores assigned to the 85 HB samples analyzed, cut-offs were identified to separate the samples into relevant subclasses. Two different cut-offs that correspond to the 33rd (0.615), and 67th percentile (0.91) have been assessed, leading to the definition of either 2 or 3 subclasses. These data together with the clinical data of 85 HB cases are given in the Table B.

Statistical Analysis of Clinical Correlations

All statistical correlations were analyzed using the discrete classification into 2 subclasses with the 67th percentile (see 3rd column of the table given in Table B).

Samples with Samples with p-values score <67th score >67th (chi- Characteristics percentile percentile square test) Previous C1/02 52/5   2/26 1.0739e−14 classification Gender Male/Female 28/29  7/21 0.03368 PRETEXT.stage 30/25 11/15 0.30367 I-II/III-IV Distant Metastasis 45/12 15/13 0.015808 No/Yes Vascular invasion 38/17 11/17 0.0090345 No/Yes Multifocality No/Yes 38/18 15/13 0.20088 Histology 34/22 16/22 0.75303 Epithelial/Mesenchymal β-catenin mutation  8/45  8/16 0.067697 No/Yes Main epithelial 49/7  5/21 2.33206e−9 component Fetal/Other*
*Other = embryonal, macrotrabecular, crowded fetal

The best correlation of the discrete classification was observed with the previous classification into C1 and C2 classes, followed by the main epithelial histological component. The correlation with patients' survival is also excellent, as shown by using the Kaplan-Meier estimates and the log-rank test. Illustrative Kaplan-Meier curves are given in FIG. 11 for specific cancer-related survival, using different percentiles to classify the tumors.

In conclusion, this study shows that the discretization method allows to classify hepatoblastoma as efficiently as the previously described method.

A similar approach was therefore applied to the analysis of hepatocellular carcinoma.

Analysis of 114 Hepatocellular Carcinomas (HC)

RNA Preparation

RNA was extracted by using either Trizol, RNeasy kit (QIAGEN) or miRvana kit (Ambion), then quantified and quality-checked by Agilent technology.

For each cDNA preparation, 1 μg of RNA was diluted at the final concentration of 100 ng/μl, and reverse transcribed with the Superscript RT kit (Invitrogen) following the manufacturer's protocol. Random primers were added at the final concentration of 30 ng/μl and the final volume was 20 μl. The cDNA was diluted 1:25, and 5 μl were used for each qPCR reaction. We added 5 μl of 2×Sybr Green Master mix or the Taqman Master mix (Applied Biosystems) and specific primers (and probes when using Taqman chemistry) at the concentration indicated by the manufacturer. Each reaction was performed in triplicate. qPCR reactions were run on the Applied Biosystems 7900HT Fast Real-Time PCR System with a 384-well thermo-block, and the conditions were the following:

2 min at 50° C. to activate Uracil-N-glycosylase (UNG)-mediated erase of aspecific reaction (omit if using the Taqman approach)

10 min at 95° C. to activate the polymerase and inactivate the UNG

40 cycles:

15 sec at 95° C. denaturation step

1 min at 60° C. annealing and extension

Final dissociation step to verify amplicon specificity (omit if using the Taqman approach)

Quantitative PCR

Real time RT-PCR was performed for 16 genes on 114 HCC samples using two different technologies:

Sybr Green as described above for hepatoblastoma (26 samples).

Taqman methodology (88 samples) using primers and probes designed and publicly released by Applied Biosystems company.

Examples

AFP forward primer: GCCAGTGCTGCACTTCTTCA AFP reverse primer: TGTTTCATCCACCACCAAGCT AFP Taqman probe: ATGCCAACAGGAGGCCATGCTTCA RHOT2 forward primer: CCCAGCACCACCATCTTGAC RHOT2 reverse primer: CCAGAAGGAAGAGGGATGCA RHOT2 Taqman probe: CAGCTCGCCACCATGGCCG

Each reaction was performed in triplicate for Sybr Green protocol and in duplicate for the taqman protocol. qPCR reactions were run on the Applied Biosystems 7900HT Fast Real-Time PCR System with a 384-well thermo-block.

Raw data for each gene were normalized to the expression of the ROTH2 gene, providing the deltaCt values that were then used for tumor classification into subclasses using the discretization method.

The normalized qPCR values (deltaCt) of the 16 genes in 26 HCC samples analyzed by the Sybr Green approach is given in Table C. The deltaCt values for 88 HCCs analyzed by the Taqman approach are given in Table D.

Analysis of qPCR Data.

The −deltaCt values for each gene in each sample was used. The cut-offs (or thresholds) selected for each gene using the Taqman method or the SybrGreen method are as follows:

Table E of cut-offs for discretization values Gene name Cut-off for Taqman Cut-off for SybrGreen AFP −1.2634010 −2.3753035 ALDH2 4.014143 5.314302 APCS 5.6142907 6.399079 APOC4 −0.7963158 4.656336 AQP9 4.2836011 5.446966 BUB1 −1.2736579 −3.634476 C1S 6.3514679 6.240002 CYP2E1 6.9562419 5.829384 DLG7 −2.335694 −4.614352 DUSP9 −7.979559 −1.8626715 E2F5 −0.4400218 −1.367846 GHR 1.0832632 1.169362 HPD 6.7480328 6.736329 IGSF1 −4.8417785 7.6653982 NLE −1.6167268 −1.82226 RPL10A 6.2483056 5.731897

For the sample, the relative expression value is determined for each gene of the set of profiled genes. Each value is compared to the cut-off for the corresponding gene and is then discretized as a result of its position with respect to said cut-off.

The next step consisted in assigning a score to each sample as follows:

1—the average of the “discretized” values of the 8 proliferation-related genes was determined. The 8 proliferation-related genes are the following: AFP, BUB1, DLG7, DUSP9, E2F5, IGSF1, NLE, and RPL10A.

2—the average of the “discretized” values of the 8 differentiation-related genes was determined. The 8 differentiation-related genes are the following: ALDH2, APCS, APOC4, AQP9, C1S, CYP2E1, GHR, and HPD.

3—The score for each sample was determined as the ratio between the to average of proliferation-related genes and the average of differentiation-related genes.

According to this calculation, a score of 2 is the theoretical maximal score for highly proliferating and poorly differentiated tumors, whereas well differentiated and slowly proliferating tumors will have a theoretical minimal score of 0.5.

Based on the scores assigned to the 114 samples analyzed, cut-offs are identified to separate the samples into relevant subclasses. Three different cut-offs that correspond to the 30rd (0.66), 50th (0.8125) and 67th percentile (0.925) have been assessed, leading to 4 different classification methods.

TABLE F of discretized values for 114 HCCs using 3 different thresholds and 4 combinations Method 1 3-class: (1): <q30 Method 2 Method 3 Method 4 (2): q30 2-class: 2-class: 2-class: Over- Follow- q67; (1): <q30 (1): <q67 (1): <q50; all.survi- up Sample score (3): >g67 (2): >q30 (2): >q67 (2): >q50 val (years) HC 001 0.6875 2 2 1 1 1 0.07 HC 003 0.6875 2 2 1 1 1 3.33 HC 004 0.7272727 2 2 1 1 0 11.48 HC 006 0.8125 2 2 1 2 1 1.25 HC 007 1.4545455 3 2 2 2 1 1.5 HC 008 1.0769231 3 2 2 2 1 8.48 HC 009 1.75 3 2 2 2 1 0.02 HC 010 1.5 3 2 2 2 1 0.95 HC 011 0.6428571 1 1 1 1 0 12.2 HC 012 0.5714286 1 1 1 1 1 0.05 HC 014 0.625 1 1 1 1 1 1 HC 015 1.6 3 2 2 2 1 1.22 HC 017 1.875 3 2 2 2 0 10.96 HC 018 1.5 3 2 2 2 1 0.39 HC 020 0.7857143 2 2 1 1 0 15.4 HC 021 1.5555556 3 2 2 2 1 0.7 HC 022 0.5625 1 1 1 1 0 11.5 HC 023 0.5 1 1 1 1 0 11.93 HC 025 0.7142857 2 2 1 1 1 15.87 HC 026 0.7142857 2 2 1 1 1 0.83 HC 027 0.8125 2 2 1 2 1 0.1 HC 028 1 3 2 2 2 1 0.1 HC 030 1 3 2 2 2 1 12.4 HC 032 0.7857143 2 2 1 1 1 0.66 HC 034 0.625 1 1 1 1 0 15.7 HC 037 0.5714286 1 1 1 1 1 0.2 HC 038 1.0769231 3 2 2 2 1 1.12 HC 041 0.8666667 2 2 1 2 1 7.44 HC 042 0.8791209 2 2 1 2 0 10.58 HC 043 0.5 1 1 1 1 0 10.9 HC 052 1.3333333 3 2 2 2 NA 0.25 HC 058 1.875 3 2 2 2 0 8.3 HC 060 1 3 2 2 2 NA NA HC 064 0.8666667 2 2 1 2 1 5.25 HC 066 0.7142857 2 2 1 1 0 8.93 HC 101 0.9230769 2 2 1 2 0 2.5 HC 102 1.625 3 2 2 2 0 0.1 HC 103 0.75 2 2 1 1 0 1.82 HC 104 0.8666667 2 2 1 2 0 2.1 HC 105 1.4545455 3 2 2 2 0 0.56 HC 106 0.5 1 1 1 1 0 2 HC 107 0.8571429 2 2 1 2 0 1.75 HC 108 1 3 2 2 2 0 1.62 HC 109 0.5 1 1 1 1 0 1.3 HC 110 0.6923077 2 2 1 1 0 1.95 HC 111 1.1818182 3 2 2 2 1 0.7 HC 112 0.8666667 2 2 1 2 0 1.48 HC 113 1.1 3 2 2 2 1 1 HC 114 0.6666667 2 2 1 1 0 0.44 HC 115 0.875 2 2 1 2 0 0.75 HC 116 0.9333333 3 2 2 2 0 0.69 HC 117 0.6 1 1 1 1 0 1.2 HC 118 0.5 1 1 1 1 0 0.93 HC 119 0.8461538 2 2 1 2 0 1.2 HC 120 1 3 2 2 2 0 0.82 HC 121 0.9285714 3 2 2 2 0 0.6 HC 122 0.6666667 2 2 1 1 0 0.75 HC 123 1 3 2 2 2 0 0.8 HC 124 0.7857143 2 2 1 1 0 0.52 HC 125 0.8181818 2 2 1 2 0 0.9 HC 126 0.8125 2 2 1 2 0 0.42 HC 127 1.6 3 2 2 2 0 0.25 HC 128 0.6095238 1 1 1 1 0 0.44 HC 129 1 3 2 2 2 1 0.15 HC 130 1.7777778 3 2 2 2 0 0.14 HC 131 0.5625 1 1 1 1 0 0.26 HC 137 1.2222222 3 2 2 2 0 5.67 HC 138 0.75 2 2 1 1 0 5.58 HC 139 1.3333333 3 2 2 2 0 6 HC 140 0.5714286 1 1 1 1 0 4.17 HC 141 0.6153846 1 1 1 1 0 5.08 HC 142 0.8888889 2 2 1 2 1 4.08 HC 143 1.375 3 2 2 2 0 2.83 HC 144 0.6153846 1 1 1 1 0 6 HC 145 0.8 2 2 1 1 0 5.58 HC 146 0.9 2 2 1 2 0 4.33 HC 147 0.6666667 2 2 1 1 0 3.83 HC 148 1.1 3 2 2 2 0 3.08 HC 149 1.2222222 3 2 2 2 1 3.42 HC 150 0.6666667 2 2 1 1 0 5.42 HC 151 0.6153846 1 1 1 1 0 2.25 HC 152 0.6428571 1 1 1 1 1 3.67 HC 153 0.6923077 2 2 1 1 1 4.83 HC 154 1.375 3 2 2 2 1 2.21 HC 155 0.8181818 2 2 1 2 0 4.1 HC 156 1.4 3 2 2 2 1 2.31 HC 157 1 3 2 2 2 1 3.59 HC 159 0.7272727 2 2 1 1 1 2.42 HC 161 0.6 1 1 1 1 0 4.47 HC 162 1.1111111 3 2 2 2 0 3.49 HC 163 0.6 1 1 1 1 1 2.21 HC 164 0.6428571 1 1 1 1 0 4.54 HC 165 0.6428571 1 1 1 1 0 4.72 HC 168 0.6 1 1 1 1 0 6 HC 169 0.6 1 1 1 1 1 2.78 HC 170 0.5625 1 1 1 1 0 5.29 HC 171 0.8181818 2 2 1 2 0 4.57 HC 172 0.8333333 2 2 1 2 0 3.9 HC 173 0.6428571 1 1 1 1 0 4.21 HC 176 0.6428571 1 1 1 1 0 4.57 HC 177 0.6666667 2 2 1 1 0 5.42 HC 178 0.7142857 2 2 1 1 0 2.5 HC 179 0.8181818 2 2 1 2 0 5.17 HC 180 0.8571429 2 2 1 2 1 3.58 HC 181 1 3 2 2 2 0 6.83 HC 182 0.5625 1 1 1 1 0 3.5 HC 183 0.7333333 2 2 1 1 1 4.08 HC 184 0.9230769 2 2 1 2 1 2.08 HC 185 0.7692308 2 2 1 1 0 2.25 HC 186 0.9285714 3 2 2 2 1 2.17 HC 187 0.6428571 1 1 1 1 0 7.67 HC 188 0.7142857 2 2 1 1 0 4.67 HC 189 0.8666667 2 2 1 2 1 3.25 HC 190 0.7619048 2 2 1 1 0 5.58

Samples were separated into the corresponding subgroups, and subsequent analysis was carried out using the 4 classification methods. Survival for each group was determined using the Kaplan-Meier estimates and the log-rank test.

Statistical Analysis of Clinical Correlations with the Subclasses for 114 HCCs

A complete table with all clinical and pathological data collected for 114 HCC patients is given in Table G. The different parameters are represented as follows:

TABLE H Clinical and pathological parameters and molecular classification of 114 HB cases. Characteristics Etiology†* Alcohol   40 (36%) HCV   26 (23%) HBV   23 (20%) Hemochromatosis   6 (5%) NASH   6 (5%) Unknown   23 (20%) Treatment (SR, OLT) 93/21 Chronic viral hepatitist†   46 (41%) Liver cirrhosis†   44 (48%) Tumor characteristics Macrovascular invasion†   20 (25%) Microvascular invasion†   47 (50%) Mean tumor size, cm (range)†  7.9 (1.5-22) Multifocality   46 (48%) Histology: Edmonson Tumor grade† (1/2/3/4) 7/35/47/5 OMS Tumor differentiation (W/M/P) 51/55/6 Classification with 16-genes by discretization 40th Percentile (C1/C2) 30/84 50th Percentile (C1/C2) 55/59 67th Percentile (C1/C2) 77/37 Mean follow-up, months (range) 43.6 (0.26-146) Tumor recurrence†   43 (40%) Alive/DOD† 75/38
Abbreviations:

HCV, hepatitis C virus;

HBV, hepatitis B virus;

NASH, Nonalcoholic steatohepatitis;

SR, surgical resection;

OLT, orthotopic liver transplantation;

W, well differentiated;

M, moderately differentiated;

P, poorly differentiated;

NA, not available;

DOD, dead of cancer.

*12 cases have more than one etiological agent and data were not available for 2 Gases.

†Data were not available for all cases. Percentages were deduced from available data.

In a second step, the intrinsic parameters of the tumors correlated with patients' survival were analyzed. In this series of tumors, only tumor grade (Edmonson) and vascular invasion were significantly correlated with survival.

TABLE I Summary of the clinical variables associated to overall survival (Kaplan-Meier curves and log-rank test). This Table does not take into account the molecular classification N. N. patients Log Variable patients Log rank With PH rank Edmonson Tumor grade 94 0.028 73 0.032 (1-2/3-4) Tumor diff. OMS 111 0.406 90 0.647 (Well/Moderate-poorly duff.) High proliferation: >10 45 0.054 34 0.402 Mitosis in 10 fields 40× (N/Y) Macrovascular Invasion 79 0.001 59 0.010 (N/Y) Microvascular Invasion 92 0.007 72 0.050 (N/Y) Tumor size≧10 cm 113 0.298 92 0.314

Classification by Discretization of Continuous Values

The clinico-pathological parameters were compared between the tumor groups using student's t test and chi-square test. Survival was analyzed by using Kaplan-Meier curves and log rank test. A special attention was given to the classification with the 67th percentile. Follow-up was closed at 146 months for overall survival (OS) and at 48 months for disease-free survival (DFS).

TABLE J Association of 16-gene classification by discretization with clinical and pathological data (chi-square test). Abbreviations: P33, 33rd percentile, P50 50th percentile and P67, 67th percentile. p-value P67 Variable P33 P50 P67 C1 C2 comments Edmonson Tumor 0.006 <0.001 <0.001 38/27 4/25 20 cases with grade: grade 1 missing values and 2 (well differentiated) vs. 3 and 4 (moderately and poorly diff.) Tumor 0.006 0.001 <0.001 45/32 6/29 2 cases with missing differentiation values OMS (Well/Moderate- versus poorly differentiated) High proliferation: 0.021 0.001 0.001 22/7   4/12 >10 mitosis in 10 fields 40× (N/Y) Macrovascular 0.097 0.033 0.008 44/8  16/12 The cases defined as Invasion (N/Y) possible are considered negative. Microvascular 0.071 0.001 0.009 37/26  9/21 The cases defined as Invasion (N/Y) possible are considered negative. Tumor size ns ns 0.015 57/20 19/18 Different cut-offs </≧10 cm assessed: 2, 3, 5 and 10 cm Multifocality (N/Y) ns ns ns 35/30 15/16 Macronodules of ns ns ns 24/9  12/4  regeneration Norm Liver A0F0- ns ns ns 48/17 27/7  A0F1 Cirrhosis AXF4 ns ns ns 31/29 17/15 (N/Y) Score METAVIR 0.053 0.044 ns 19/32  5/20 Activity >0 (N/Y) Score METAVIR ns 0.20 ns 31/20 15/10 Activity >1 (N/Y) Score METAVIR 0.041 ns ns  5/48  2/27 Fibrosis >0 (N/Y) Score METAVIR ns ns ns 19/35  7/22 Fibrosis >1 (N/Y) Score METAVIR ns ns ns 24/30  8/21 Fibrosis >2 (N/Y) Score METAVIR ns ns ns 26/28 15/14 Fibrosis >3 (N/Y) Chronic viral 0.047 ns ns 48/29 18/17 hepatitis (N/Y) HBV (N/Y) 0.075 ns ns 62/15 27/8  HCV (N/Y) ns ns ns 61/16 25/10 Alcohol (N/Y) ns ns ns 47/30 25/10 Recurrence (N/Y) ns ns ns 41/32 24/11 HCC034 and HCC030 censored Survival (N/Y) 0.050 0.023 0.031 56/21 19/17 HCC025 and HCC030 censored DFS (N/Y) ns ns ns 35/42 15/21 HCC025 and HCC030 censored

In conclusion, these data show significant correlations between molecular classification using the 3 methods and the following parameters: Tumor grade (Edmonson), tumor differentiation (OMS), proliferation rate, vascular invasion and survival. In contrast, the classifications were not correlated with etiological factors (viral hepatitis, alcohol, etc. . . . ), with the state of the disease in adjacent, non tumoral livers or with tumor recurrence.

The data suggest that classification using the 67th percentile seems to be the most adequate and is strongly recommended to classify HCCs.

Multivariate Analysis

To further determine the efficiency of the molecular classification using the 67th percentile, we performed multivariate analysis with the Cox regression test on two sets of patients for which all data were available:

91 patients that received either surgical resection or orthoptic liver transplantation (OLT)

71 patients that received surgical resection.

Different variables associated to survival in the clinical settings have been included in the multivariate analysis: 1) Edmonson grade, 2) microvascular invasion and 3) Molecular classification using the 67th percentile.

TABLE K Multivariate test (Cox regression). N patients variable HR 95% CI p-value 91 Molec classsif (p67) 2.534 (1.214-5.289) 0.016 (surgical Edmonson Tumor grade 1.690 (0.747-3.823) 0.205 resections 1-2/3-4) and OLT) Microvascular Invasion 2.451 (1.105-5435) 0.024 (N/Y) 71 Molec classsif (p67) 2.646 (1.1156.278) 0.032 (only Edmonson Tumor grade (1- 2.697 (1.103-6.592) 0.026 surgical 213-4) resections) Microvascular Invasion 1.681 (0.648-4.359) 0.282 (N/Y)

Correlation of the Molecular Classifications with Survival

For overall survival (OS) and disease-free survival (DFS), we compared the efficiency of the 3 methods of discretization that separate the samples into 2 subclasses. Independent studies were made for patients that received surgical resection and for patients that received orthoptic liver transplantation (OLT). The ability of the 16-gene signature to discriminate between recurrent and non-recurrent tumors was also assessed.

Table L Summary of survival analysis using Kaplan-Meier curves and log-rank test Analysis N. patients Classif. method Log rank OS 113 P33 0.037 113 P50 0.005 113 P67 0.002 DFS 113 P33 0.078 113 P50 0.019 113 P67 0.072 recurrence 108 p33* 0.134* 108 p50* 0.115* 108 P67 1.000 Analysis of 92 cases that received surgical resection OS 92 P33 0.032 92 P50 0.009 92 P67 0.013 DFS 92 P40 ns 92 P50 ns 92 P67 ns recurrence 88 P33 ns 88 P50 ns 88 P67 ns
Abbreviations: OS, overall survival; DFS, disease free survival

*There is a trend but it is not significant and it is lost in the P60 analysis

The different analyses are illustrated in the Kaplan-Meier plots shown in FIG. 12. The discretization method of classification showed the same efficiency in the analysis of tumors obtained either from surgical resection (also called partial hepatectomy, PH) or from orthotopic liver transplantation (OLT), showing that the clinical management of the tumor had no impact on the classification.

In conclusion, the method described herein is able to classify HCC cases according to tumor grade and patient's survival, and represents a powerful tool at diagnosis to stratify the tumors according to the prognosis, and for further clinical management of HCC. In particular, it may be an excellent tool for the decision of orthotopic liver transplantation, since the criteria used currently are limited and often poorly informative of the outcome.

Protocol for Applying the Method to a New Sample

The following protocol is designed according to the invention:

1—extract total RNA from the tumor specimen using well established technologies.

2—synthesize cDNA synthesis (suggested conditions: 1 μg RNA and 300 ng of random hexamers for a 20 μl-reaction)

3—amplify the selected genes said genes being in equal number of each of the groups defined as overexpressed proliferation-related genes group and downregulated differentiation-related genes group (profiled genes within the group of 2 to 16 genes) and the reference gene (invariant gene) such as for example the RHOT2 gene 1:5 cDNA dilution, using either Taqman or SybrGreen qPCR technology.

4—determine the Delta Ct (DCt) value for each gene

5—compare the value with the threshold of reference (for HB or for HC) in order to assign a discretized value of “1” or “2”.

5—determine the average of discretized values in each group, i.e., for the selected proliferation-related genes (up to 8) separately for and the selected differentiation-related genes (up to 8) and determine the ratio of these 2 average values which is the score of the sample.

6—compare the result with the reference scores corresponding to the following cut-offs:

C1

|30rd=0.6667

|50th=0.8125

|67th=0.925

C2

Example

For patient X having an HC tumor a Taqman qPCR is performed.

Step one: assignment of discretized values to each selected gene among proliferation-related genes and differentiation-related genes.

Example

The DCt of AFP is −4.0523

The cut-off for AFP for qPCR using Taqman technology is −1.2634010 Given that −4.0523 is lower than the cut-off, the assigned discretized value is 2.

Step two: Determination of the average of discretized values for the 2 sets of 8 genes:

AFP=2; BUB1=1; DLG7=2; DUSP9=2; E2F5=2; IGSF1=1; NLE=2; RPL10A=1;

Average of Proliferation-Related Genes: (2+1+2+2+2+1+2+1)/8=1.625

ALDH2=1; APCS=1; APOC4=1; AQP9=1; C1S=2; CYP2E1=2; GHR=1; HPD=2;

Average of Differentiation-Related Genes: (1+1+1+1+2+2+1+2)/8=1.375

Step Three: calculate the ratio proliferation/differentiation score.

In this example: 1.625/1.375=1.18182

Step 4: compare the result with the reference scores:

C1

|30rd percentile=0.6667

|50th percentile=0.8125

|67th percentile=0.925

C2

Classification based on the value of the ratio=1.18182.

As the value is above the 67th percentile, the assigned class is C2.

TABLE A id AFP ALDH2 APCS AP0C4 AQP9 BUB1 C1S CYP2E1 HB1 −7.684892 −4.592702 −0.660189 −2.651319 -4.194894 -1.068025 -1.394659 -3.334692 HB100 −7.682724 −3.849128 −0.372566 0.297278 −0.305738 0.65983 −2.572264 −7.352142 HB101 1.801478 −7.157316 −1.166513 −4.924476 −8.067838 6.222865 −5.284734 −11.757699 HB102 −7.761115 −5.696697 −1.044129 −2.374592 −3.447046 2.724363 −3.657616 −5.769417 HB103 2.908026 −2.580629 −2.748625 −2.55635 1.480624 3.891875 −2.819372 0.454623 HB106 0.294848 −7.534485 −1.424535 −5.377043 −7.886612 4.855797 −6.80698 −11.496242 HB107 0.719866 −6.546079 −9.18522 −3.425075 −6.189664 3.901806 −5.609115 −10.6711555 HB11 1.492805 −3.560021 −5.094387 −1.031623 −8.42849 2.086834 −6.166353 −9.043371 HB112 4.155252 −6.486961 −0.154814 −4.48155 −5.634596 3.762347 −7.88579 −8.960815 HB114 6.2971 −3.966456 5.02266 0.604275 3.037682 4.23408 −5.29691 −0.313326 HB118 0.318307 −4.311795 −5.146409 −3.787568 −5.428442 2.329959 −5.284827 −7.342423 HB121 −0.971033 −6.879043 −8.355819 −4.679393 −6.361435 2.329708 −6.559457 −8.87105 HB122 2.188721 −6.220957 −7.7399 −3.410743 −5.745306 3.309004 −6.327656 −8.906339 HB125 2.929931 −4.053616 −4.882212 −2.32494 −3.352398 5.067815 −4.255762 −7.887455 HB126 2.458273 −5.577951 −6.518289 −3.182407 −5.243351 5.270089 −5.814672 −8.188307 HB129 −4.930877 −2.124281 −0.744262 1.154663 −0.846572 0.421372 −2.925458 −4.708874 HB130 −4.86199 −1.139837 −1.398588 0.115559 −1.313951 1.669543 −2.37235 0.175598 HB131 5.545406 −1.714367 −1.045683 2.628822 1.903853 1.972112 −2.306818 0.069456 HB132 2.654369 −3.71955 −6.543987 −3.876868 −4.7099 4.043489 −4.801651 −7.725089 HB136 5.005516 −3.234557 −4.827283 2.471208 −0.502385 −1.945351 −4.324749 −4.844765 HB140 2.835457 −7.041546 −6.88604 −5.561912 −5.089682 4.140594 −6.023758 −10.477228 HB142 5.200474 −4.919616 2.416807 2.058522 −3.396171 1.380591 −5.965126 1.196438 HB145 3.58286 −5.186236 −5.18731 NA −5.118895 5.58416 −5.786933 −7.880334 HB146 −1.290056 −5.422341 −5.973879 −3.869993 −5.908024 0.982626 −4.124487 −8.751883 HB147 −9.442257 −3.655303 −0.362122 1.179633 −2.349782 −1.51351 −2.756099 0.30832 HB148 −3.566401 −5.382548 −6.721533 −2.380348 −6.951359 1.183916 −4.188648 −7.101147 HB150 2.356994 −5.56181 −5.496186 −4.45536 −5.603247 5.136577 −5.435261 −8.522001 HB153 −2.086302 -4.364035 -4.049735 -1.1908 -4.342186 2.437297 -6.055092 -7.522683 HB155 −1.951256 −5.140738 −7.17357 −0.801318 4.538929 4.038538 −5.939438 3.058475 HB156 −6.523604 −4.658012 −5.112322 −1.499462 −1.13031 1.970226 −4.763811 −8.138508 HB157 −8.747252 −3.193287 −0.914511 0.563787 −0.139273 0.648195 −3.089302 −2.404646 HB160 4.40621 −0.878277 −2.381785 −1.9527 0.770799 4.516203 −2.89522 1.197611 HB162 −1.127062 −5.142195 −6.564426 −2.432348 −5.179601 3.27157 −4.959578 −9.351464 HB165 −1.015428 −1.578048 −1.612095 −1.677494 1.921123 −0.416058 −4.579384 −0.458984 HB167 −7.323435 −5.692388 −6.461153 −2.470512 −4.912208 −0.369976 −4.949694 −10.583324 HB170 −0.980072 −5.786627 −7.265156 −3.690367 −5.952908 1.548967 −6.61768 −8.574004 HB171 2.310988 −5.687635 −7.127181 −3.794631 −5.898635 2.05689 −6.420469 −8.856566 HB172 4.547243 −0.385469 −1.804453 −1.833478 2.11442 4.373205 −3.929151 1.277285 HB173 1.889759 −5.184791 −4.471618 −2.235657 −5.743057 2.116789 −4.966413 −7.319851 HB175 −2.0436 −6.05152 −8.152949 −2.996302 −3.829205 3.036838 −5.151913 −9.108766 HB184 −6.561121 −2.895788 −5.35813 −1.653786 0.293844 −0.082754 −3.084271 −3.362889 HB20 4.752153 −4.811256 −5.712608 −2.133951 −5.361771 5.572378 −4.283688 −8.390209 HB28 −4.001793 −4.719296 −7.514733 −2.385516 −3.869707 0.599685 −5.187286 −9.373678 HB3 0.027392 −4.565046 −4.462833 −2.255273 −4.14636 4.676108 −5.373064 −6.610781 HB33 −7.497741 −3.066759 −5.881277 0.250334 0.950966 0.500246 −3.829096 −6.510795 HB39 −8.613403 −3.166427 3.421734 1.699859 −0.944463 −0.146929 −1.480822 −0.727464 HB48 −4.768603 −3.632136 −4.882397 −2.170561 −4.965403 1.366439 −3.944489 −9.061667 HB49 1.818606 −5.933777 −5.948111 −4.936781 −5.434931 4.576628 −5.318794 −9.381172 HB5 −2.282703 −6.147963 −7.059143 −4.107155 −7.593099 2.501017 −6.573836 −9.813634 HB54 1.132255 −4.844075 −5.655802 −2.937193 −4.595442 3.040468 −4.999207 −8.199672 HB59 1.334928 −6.792009 −7.221196 −5.590302 −6.300828 1.42553 −5.648808 −9.279234 HB6 −1.610623 −7.099329 −7.979286 −5.729452 −5.2647225 2.920021 −5.482511 −10.151809 HB60 −0.594337 −5.206398 −6.67766 −1.663871 −2.889326 3.97632 −5.504179 −6.743858 HB61 −5.058775 −6.113525 −5.991888 −3.527984 −5.387419 3.269827 −6.119246 −8.943929 HB62 −1.989342 −4.487171 −6.502588 −0.923844 −4.712471 3.449967 −4.22945 −7.087853 HB63 −0.891056 −4.153057 −5.680458 −2.637115 −5.710062 4.49543 −2.939154 −9.095241 HB65 3.025127 −4.346225 −5.338104 −1.175748 −1.226393 −0.613979 −5.196916 −4.645702 HB66 -1.861761 -4.166485 -5.897819 -2.09279 -3.003258 4.774807 -4.585607 -6.839392 HB68 −4.313608 −6.550704 −6.762513 −3.66757 −5.982654 4.060667 −5.956246 −8.393607 HB69 −1.820363 −9.245314333 −8.965648 −7.384871667 −9.430164667 −2.026701667 −8.961309 −12.31658 HB7 1.334084 −4.488213 −5.853708 −2.13753 −5.142938 4.894117 −4.082335 −8.118103 HB70 2.021391 −5.678476 −7.496267 −5.781771 −4.346458 2.174971 −7.066038 −8.392057 HB72 −11.99570467 3.978023333 −1.371737333 −2.543168667 −6.278723667 −5.504070333 −7.162789667 −8.103601333 HB73 −10.69629133 −8.263771333 −4.869197667 −2.900671333 −5.802080667 −5.324255333 −8.090371 −9.754354333 HB74F 3.831288 −7.73216 −4.940396 −6.3439 −6.355995 6.130615 −5.584023 −10.472842 HB75 0.474553 −6.309769 −2.777247 −4.334006 −6.807299 4.545387 −5.115577 −10.418948 HB77 2.915987 −5.645872 −6.698372 −2.284956 −5.392377 4.544876 −5.559466 −8.695429 HB78 −3.945686 −2.82555 −2.986284 −1.790335 −0.938738 4.523136 −2.620165 −5.945013 HB79 −0.781193 −5.652768 −5.454157 −3.953162 −5.051444 0.254305 −5.44242 −9.05667 HB8 −6.696169 −3.108913 0.498461 1.361801 −3.322642 0.055848 −0.348492 −1.877119 HB80 −8.8331005 −4.713883 −2.9124615 −2.810437 −0.838727 −0.7226515 −2.5925445 −5.408417 HB81 −4.851198667 −10.55296467 −10.55292033 −7.621321667 −10.19195633 −2.962795333 −10.17992067 −12.72629433 HB82 −1.942166 −5.620028 −5.739178 −3.972123 −6.520482 0.934055 −3.737063 −8.932744 HB83 −4.169107 −9.660034667 −9.382586667 −8.05219 −10.951863 −3.521245667 −10.12345167 −9.850559667 HB86 −6.283735 −5.287677 0.896101 −1.494853 −2.934412 −0.46896 −2.879366 −5.76077 HB89 2.996384 −7.323446 −7.464817 −5.120874 −5.856518 4.907738 −6.676481 −9.415603 HB9 −3.679937 −4.761778 −6.571455 −2.775269 −6.201772 2.209541 −3.895565 −8.86438 HB90 2.024206 −8.47846 −1.33932 −6.745716 −6.677122 5.899195 −8.114672 −10.459034 HB93 −4.610162 −5.583852 −5.277197 −1.990982 −2.698011 −1.085743 −4.488914 −3.388975 HB94 1.79868 −5.621254 −7.718202 −6.940586 −6.67335 3.551727 −6.54809 −8.572742 HB95 −0.444835 −5.745006 −8.404602 −5.637613 −6.396063 6.671045 −5.701559 −10.554918 HB96 −4.775396 −6.402052 −6.123253 −4.340961 −5.066688 3.365736 −6.521753 −9.090145 HB97 −6.841231 −6.21691 −6.275051 −3.638382 −3.617558 2.362203 −6.58495 −5.781372 HB98 −4.911783 −2.946932 6.478933 4.211147 0.395926 2.311268 −2.827802 0.584022 HB99 −4.551378 −1.14591 −5.549696 −1.796859 1.62906 2.600714 −2.483835 −3.848236 HB1 4.140368 −5.212318 −0.812424 1.207583 3.840983 −0.715134 −0.812792 −8.675945 HB100 4.399124 −5.749706 0.27698 1.907294 −0.113253 −2.800323 0.547899 −6.153046 HB101 7.086329 −0.641871 0.737702 −3.913751 −4.340259 7.086329 0.191689 −6.757648 HB102 7.380694 −4.303866 1.144778 0.2784 −0.284245 −2.545668 0.856607 −6.803817 HB103 5.997143 0.880421 3.697478 1.249386 −2.713306 1.392197 −0.453035 −4.535615 HB106 6.79755 −1.540745 0.77722 −4.155098 −5.747164 2.274385 0.291903 −6.637275 HB107 5.239962 −1.184244 3.145996 −1.891404 −4.433271 3.119114 −0.053334 −6.319917 HB11 3.688558 −1.412987 −0.179621 −0.149048 −1.897658 2.297186 −0.19686 −5.623341 HB112 6.035002 −2.179125 −0.998979 −3.575994 −4.671755 −0.776138 −2.252113 −7.8479 HB114 6.2971 2.615827 0.886564 0.002487 1.919397 2.50863 1.785623 −7.055851 HB118 3.935101 2.405105 2.275962 −0.451819 −4.812319 2.339813 0.486307 −5.904633 HB121 3.458157 −2.1882 1.247645 −1.155575 −5.938235 3.750147 1.867907 −5.131548 HB122 3.562777 1.229723 2.386559 −1.961029 −5.590919 2.406687 1.976893 −5.368023 HB125 5.700252 0.274642 2.864883 0.118717 −3.155289 2.138032 −0.470879 −3.478449 HB126 6.32602 0.274197 3.089709 −1.334371 −5.227705 2.726599 0.54385 −4.787822 HB129 4.474485 −3.829751 1.158283 3.025728 1.984295 −0.074354 1.326073 −5.682215 HB130 5.297728 −2.554008 2.251163 3.317556 0.885962 0.039307 1.389742 −4.829542 HB131 5.801168 2.269272 2.226921 1.235598 2.035452 5.621114 1.777334 −4.96776 HB132 8.18041 0.433104 4.507503 −0.157093 −2.441422 5.855213 2.895208 −3.579579 HB136 1.140686 0.10165 −2.336947 0.261203 0.124159 3.807218 −0.676358 −7.113232 HB140 9.015818 −0.401264 2.325356 −3.379816 −3.148068 3.156456 0.80129 −7.308986 HB142 6.203192 4.554631 3.03661 2.598877 4.150455 8.782461 1.428955 −6.630178 HB145 6.734264 1.908734 2.518779 −1.358174 −5.181668 4.610406 1.707345 −4.6775 HB146 0.991164 −0.681828 0.1227 −0.510651 −4.471483 0.777004 0.176935 −5.992209 HB147 −1.376061 −4.733546 −2.588397 1.772494 −1.944032 −2.698708 −0.565682 −7.527854 HB148 1.7033 −1.806502 −0.663069 −1.376372 −5.121145 −0.683001 −0.431826 −6.201895 HB150 5.800233 0.8436 2.758596 −1.181738 −5.492037 2.891937 0.439392 −4.69542 HB153 3.096912 −2.657862 0.449197 −0.480929 −4.261986 3.34336 1.423023 −5.963837 HB155 4.360922 −1.23259 0.752365 −3.062474 0.657144 −1.091013 0.911424 −5.964497 HB156 2.483547 −1.214228 0.687246 −1.107338 −3.806189 −1.181305 0.159847 −5.65452 HB157 0.181175 −4.1451 0.297747 1.940187 −3.850885 −1.38623 0.041349 −5.820536 HB160 6.224569 2.906158 4.403545 2.633949 −2.138569 3.355814 −0.100123 −4.568688 HB162 4.25017 −1.453283 1.117439 −0.163468 −4.733881 1.809885 −0.022627 −4.822098 HB165 −0.010488 1.837305 0.47467 −2.953007 −0.655058 −1.791164 −0.933062 −5.535221 HB167 0.509668 −1.707485 0.198742 0.269552 −4.442331 −1.197651 −0.240385 −5.755341 HB170 2.567207 1.148738 1.360144 −2.397242 −4.944439 2.424619 −0.463297 −5.539725 HB171 2.278353 1.67404 2.062277 −1.193735 −4.984552 2.19098 0.230044 −4.81411 HB172 6.060459 2.366999 3.689341 2.93017 −1.316921 2.571021 −0.153162 −3.812616 HB173 2.779999 1.921427 3.05205 −0.20919 −4.475376 0.418818 0.678606 −4.361307 HB175 4.414558 −1.623242 1.49 −0.662783 −4.684446 3.524049 1.78088 −5.173616 HB184 1.361379 −1.542307 −0.588812 1.814793 −2.048922 −0.326393 0.097971 −4.663763 HB20 9.423325 −0.34174 2.066057 −0.975735 −3.695854 4.361484 1.157495 −5.27136 HB28 1.922989 −2.304861 1.222545 −0.120436 −5.154703 −0.192738 1.819854 −5.824864 HB3 7.285685 0.65201 2.301029 −0.049158 0.117373 4.46221 1.743745 −6.911792 HB33 1.659659 −4.338262 −0.148233 1.134133 −4.625204 −2.34198 1.272614 −5.63922 HB39 2.485354 −4.927491 −1.241931 1.694781 −0.33289 −2.652634 −0.149609 −6.579218 HB48 1.583391 −3.620772 −0.089081 1.342382 −2.330218 0.686163 1.169838 −6.508074 HB49 5.652893 2.41148 3.776672 −1.220476 −5.746779 4.727596 2.190021 −4.286949 HB5 3.674234 −2.082424 0.98073 −1.943451 −6.561791 1.592167 0.449005 −6.230808 HB54 3.556268 3.982183 3.025795 −0.158057 −4.638333 3.623678 1.995039 −5.061096 HB59 5.127336 0.250753 3.459226 −2.269072 −4.727738 6.045093 1.466312 −6.48303 HB6 6.733353 −0.246309 3.812183 −2.459856 −3.728987 0.835057 2.205872 −7.208765 HB60 5.188517 2.869544 3.228365 −0.276338 −4.031974 2.026116 2.577353 −4.502382 HB61 5.827933 −5.51457 1.00606 −3.272672 −4.816797 −0.203871 0.753758 −6.140918 HB62 4.328277 0.708512 1.218963 1.021692 −3.265138 0.731519 2.223877 −5.334147 HB63 5.003075 −1.082094 0.951357 1.316553 2.000601 4.964996 1.31674 −6.741518 HB65 2.978487 −0.087486 −1.274388 0.080222 −2.417946 1.06702 −1.371523 −6.195428 HB66 8.039274 −0.423313 2.141981 −1.148424 −1.349111 −0.305017 1.586659 −5.393141 HB68 7.010986 −0.530541 2.520261 0.232431 −1.779051 −0.603113 2.342104 −4.959414 HB69 3.071106 −0.626059667 3.421015 −5.118794333 −6.824055667 11.819556 −0.603036 −2.847600667 HB7 8.076437 −0.833011 1.354912 −0.884629 −2.106592 2.978739 2.384133 −5.458546 HB70 4.083519 3.896364 2.616204 −3.614294 −6.063097 2.060379 1.506083 −4.669554 HB72 −1.688566667 −8.976227 −1.809694 −1.750672 −3.40203 −6.090071333 −2.505424 −5.054027 HB73 −2.068555667 −9.537516 −1.965151 −0.544775 −5.542041333 −7.013002667 −3.078154667 −5.580986333 HB74F 8.986048 0.497828 4.585503 −2.916191 −3.041943 7.759608 1.654283 −6.380865 HB75 7.231393 −2.411839 0.378995 −1.925637 −5.055106 2.61456 1.017432 −5.77539 HB77 9.66177 −0.139299 2.727198 −1.675013 −4.079932 2.793758 2.146337 −4.964228 HB78 5.293419 −0.185629 1.735594 0.020191 −3.984125 −2.010153 −0.114956 −3.94071 HB79 1.90306 1.145681 1.319285 −1.978228 −5.757335 0.01942 −0.194167 −5.016158 HB8 1.950257 −4.043236 −1.814636 2.280516 1.100353 0.314694 0.29834 −7.823095 HB80 2.660644 −4.9166885 −0.374031 0.675995 −0.4253495 −4.2048885 −0.8782055 −7.919531 HB81 2.155925333 −5.738363 0.932455333 −5.565798 −8.171378 −1.999123333 −2.092100667 −4.795482 HB82 1.47049 −3.938165 −0.549544 −1.023595 −3.267403 8.008069 0.067941 −7.635394 HB83 2.492243 −4.003930333 4.737920667 −4.561133333 −6.966227667 −0.028684333 −0.855054667 1.789090833 HB86 3.219092 −5.894534 −0.496662 0.35847 −0.121981 −2.31061 0.046472 −8.510995 HB89 8.255339 1.284916 3.638735 −2.665258 −5.177704 3.273649 1.279167 −5.898171 HB9 4.940411 −1.989636 0.700504 −0.698988 −3.255601 2.609339 1.300875 −6.54224 HB90 6.54891 1.104162 1.408459 −5.754423 −7.507485 4.45026 1.52717 −6.250036 HB93 3.902565 −7.483471 −0.488108 0.969648 −1.415501 −1.818147 −0.829773 −7.824402 HB94 8.669386 −1.132305 0.490788 8.498726 −6.819645 7.800646 −0.149162 −5.793072 HB95 6.921267 −1.620869 2.726241 −2.193777 −5.454765 1.364738 0.279802 −5.172451 HB96 6.685021 −0.591271 1.973021 −4.924202 −4.91283 1.722505 1.829525 −5.638435 HB97 6.474525 −5.800537 1.05047 −0.911789 −4.571465 −4.308964 −0.87035 −6.60257 HB98 6.837198 −2.065483 2.482301 1.17723 −0.98407 −0.701098 1.175939 −5.166874 HB99 6.353711 −4.201828 1.467552 1.703655 −0.109186 −0.822266 1.226265 −3.572067

TABLE B 67th percentile- percentile- related related previous 16 gene tumor score score score based clas- AFP at diagnosis Treatment PRETEXT ID (ratio) (2-classes) (3-classes) sification Gender Age months ng/mL treatment protocol stage HB122 0.5 1 1 C1 M 10 8000 Y H I HB126 0.5 1 1 C1 F 12 153840 Y S II HB145 0.5 1 1 C1 M 7 56000 Y S II HB150 0.5 1 1 C1 F 5 82000 Y S III HB175 0.5 1 1 C1 M 9 220000 Y S I HB20 0.5 1 1 C1 F 50 880 Y S II HB49 0.5 1 1 C1 F 15 11000 Y S II HB54 0.5 1 1 C1 M 10 180 N N I HB70 0.5 1 1 C1 F 42 812 Y S II HB77 0.5 1 1 C1 F 9 204000 Y S II HB89 0.5 1 1 C1 M 13 448 Y S I HB95 0.5 1 1 C1 M 28 1000000 Y H IV HB118 0.53333333 1 1 C1 M 17 14500 Y S NA HB132 0.53333333 1 1 C1 F 23 2078 Y NA III HB121 0.5625 1 1 C1 F 14 296000 Y S III HB140 0.5625 1 1 C1 M 3 22758 Y S II HB162 0.5625 1 1 C1 F 9 960000 Y S III HB171 0.5625 1 1 C1 F 17 300 Y S II HB173 0.5625 1 1 C1 F 27 66810 Y S I HB59 0.5625 1 1 C1 F 24 5643 Y S II HB6 0.5625 1 1 C1 M 24 320000 Y S II HB74F 0.5625 1 1 C1 M 96 325 N N I HB96 0.5625 1 1 C1 M 101 2265000 Y H IV HB60 0.57142857 1 1 C1 F 30 1990800 Y H II HB7 0.57142857 1 1 C1 M 33 45000 Y S I HB101 0.6 1 1 C1 M 42 67747 Y S III HB106 0.6 1 1 C1 F 11 320000 Y H IV HB90 0.6 1 1 C1 F 74 300 N N II HB62 0.61538462 1 2 C1 M 16 1708400 Y H IV HB107 0.625 1 2 C1 M 30 16000 Y H IV HB170 0.625 1 2 C1 M 20 123000 Y H III HB5 0.625 1 2 C1 M 84 300000 Y H III HB125 0.64285714 1 2 C1 F 15 360000 Y H IV HB75 0.66666667 1 2 C1 M 21 131000 Y S II HB9 0.66666667 1 2 C1 F 16 84000 Y NA III HB94 0.66666667 1 2 C1 M 29 1270 Y S I HB61 0.6875 1 2 C1 F 126 346000 Y NA IV HB69 0.6875 1 2 C1 M 25 1163 Y S I HB79 0.6875 1 2 C1 M 144 1200 Y S II HB3 0.69230769 1 2 C1 F 22 3192 Y S I HB66 0.69230769 1 2 C1 M 6 1000000 Y S III HB68 0.71428571 1 2 C1 F 11 119320 Y S III HB146 0.73333333 1 2 C1 F 11 NA N N NA HB155 0.75 1 2 C2 M 9 849500 Y S II HB63 0.75 1 2 C1 M 204 NA N N III HB11 0.76923077 1 2 C1 F 18 626100 Y H IV HB153 0.78571429 1 2 C1 F 27 1000000 Y H IV HB28 0.8125 1 2 C1 M 34 172500 Y NA II HB83 0.8125 1 2 C1 M 15 285 Y S II HB156 0.85714286 1 2 C2 F 2 468000 Y S III HB112 0.86666667 1 2 C1 M 36 725 Y S II HB82 0.86666667 1 2 C1 M 120 179000 N N II HB97 0.86666667 1 2 C1 F 42 700000 Y H IV HB81 0.875 1 2 C1 M 22 322197 Y H III HB103 0.9 1 2 C2 F 57 750000 Y H IV HB114 0.9 1 2 C2 F 21 8783 Y S II HB142 0.90909091 1 2 C2 F 48 605000 Y H III HB148 0.93333333 2 3 C1 M 17 200730 Y S II HB167 0.93333333 2 3 C2 M 34 1500000 Y H NA HB73 0.9375 2 3 C2 F 24 667786 Y H III HB131 1 2 3 C2 M 6 7511 Y H II HB65 1 2 3 C2 M 6 1740 N N III HB78 1 2 3 C1 M 126 376000 Y S II HB72 1.07142857 2 3 C2 F 16 1412000 Y S III HB48 1.07692308 2 3 C2 M 72 35558 Y H IV HB102 1.09090909 2 3 C2 M 41 1331000 N N II HB160 1.125 2 3 C2 M 45 342000 Y H II HB172 1.125 2 3 C2 M 50 64170 Y H II HB99 1.22222222 2 3 C2 M 72 277192 N N IV HB130 1.25 2 3 C2 F 19 1980000 Y H II HB98 1.25 2 3 C2 M 60 1285000 Y H III HB136 1.3 2 3 C2 M 6 31828 Y S III HB165 1.3 2 3 C2 M 13 18600 Y S II HB1 1.36363636 2 3 C2 F 43 3000 Y H IV HB93 1.36363636 2 3 C2 M 22 107000 Y S III HB129 1.375 2 3 C2 M 96 14000 N N I HB33 1.4 2 3 C2 M 12 765890 Y H IV HB100 1.44444444 2 3 C2 M 48 576000 N N III HB184 1.44444444 2 3 C2 M 41 912500 Y H IV HB157 1.55555556 2 3 C2 M 7 356000 Y H NA HB80 1.6 2 3 C2 M 180 37000 Y H III HB86 1.66666667 2 3 C2 M 0.08 74000 N N III HB8 1.75 2 3 C2 F 8 44610 Y NA II HB147 2 2 3 C2 F 9 2355000 Y S II HB39 2 2 3 C2 F 11 862067 Y S III Main Epi- thelial beta- tumor Distant Vascular Multi- Histol- com- catenin Follow-up Surgery Follow-up ID Metastasis invasion focality ogy ponent status (months) Outcome Type speOS (years) HB126 N N S Mx F mut 18 A R 0 1.5 HB145 N N S Mx F mut 17 A R 0 1.416666667 HB150 N N S Mx F mut 14 A R 0 1.166666667 HB175 N N M Mx F NA 6 A R 0 0.5 HB20 N N M Mx F mut 7 A R 0 0.583333333 HB49 N N S Ep F mut 42 A R 0 3.5 HB54 N N S Ep F mut 6 D R 0 0.5 HB70 N N S Ep PF mut 49 A R 0 4.083333333 HB77 N N S Ep PF mut 53 R R 0 4.416666667 HB89 N N S Ep F mut 37 A R 0 3.083333333 HB95 N N S Ep F mut 33 A R 0 2.75 HB118 Y Y M Mx F mut 32 A LT 0 2.666666667 HB132 N N S Mx F mut 121 A R 0 10.08333333 HB121 N N M Mx F mut 18 A R 0 1.5 HB140 N N S Mx F mut 22 A R 0 1.833333333 HB162 N N S Mx F mut 13 A R 0 1.083333333 HB171 N N S Ep F mut 9 A R 0 0.75 HB173 N N S Ep F NA 11 A R 0 0.916666667 HB59 N N S Ep PF mut 72 A R 0 6 HB6 N Y S Ep F mut 48 A R 0 4 HB74F N Y S Ep F mut 35 A R 0 2.916666667 HB96 N Y M Ep F mut 23 R LT 0 1.916666667 HB60 N Y S Ep F wt 63 A R 0 5.25 HB7 N Y S Mx F mut 46 A R 0 3.833333333 HB101 N N S Ep F mut 20 A R 0 1.666666667 HB106 N N S Mx F mut 25 A R 0 2.083333333 HB90 N N S Ep F mut 35 A R 0 2.916666667 HB62 N N S Mx F mut 69 A R 0 5.75 HB107 Y Y M Ep F mut 25 A LT 0 2.083333333 HB170 Y Y M Ep F wt (FAP) 15 A R 0 1.25 HB5 Y Y M Ep F mut 24 DOD R 1 2 HB125 Y N M Mx F mut 17 A LT 0 1.416666667 HB75 N Y S Mx F mut 41 A R 0 3.416666667 HB9 N N S Ep PF mut 91 A R 0 7.583333333 HB94 N N S Ep PF wt 29 A R 0 2.416666667 HB61 Y Y M Mx F mut 5 DOD R 1 0.416666667 HB69 N N S Ep PF wt 55 A R 0 4.583333333 HB79 N N M Ep M mut 39 A LT 0 3.25 HB3 N N S Ep F wt 55 A R 0 4.583333333 HB66 N N S Ep F mut 68 A R 0 5.666666667 HB68 N N S Mx E mut 52 A R 0 4.333333333 HB146 N NA S NA NA NA 1 D R 0 0.083333333 HB155 N N S Mx CF mut 8 A R 0 0.666666667 HB63 N Y M Mx F mut 96 A R 0 8 HB11 Y Y M Mx F mut 21 DOD R 1 1.75 HB153 Y N M Mx CF mut 8 A LT 0 0.666666667 HB28 N N S Ep F wt 120 A R 0 10 HB83 N N S Ep PF mut 53 A R 0 4.416666667 HB156 N N NA Ep F NA 6 A R 0 0.5 HB112 N N S Ep F wt 32 A R 0 2.666666667 HB82 N N S Ep F mut 63 A R 0 5.25 HB97 N Y M Ep F mut 30 A R 0 2.5 HB81 Y Y M Ep F mut 36 A R 0 3 HB103 Y Y M Ep M mut 9 DOD M 1 0.75 HB114 N N S Mx E mut 23 A P 0 1.916666667 HB142 Y Y S Ep NA mut 16 A R 0 1.333333333 HB148 N N S Mx F mut 11 A R 0 0.916666667 HB167 Y Y M Ep F mut 2 A R 0 0.166666667 HB73 Y Y S Ep E mut 16 DOD R 1 1.333333333 HB131 Y N S Ep E wt 1 DOD R 1 0.083333333 HB65 N N M Mx E wt 2 DOD R 1 0.166666667 HB78 N Y M Ep CF wt 32 A R 0 2.666666667 HB72 Y Y M Mx E mut 9.5 DOD R 1 0.791666667 HB48 N Y M Ep CF mut 9 DOD R 1 0.75 HB102 N N S Ep CF mut 4 D B 0 0.333333333 HB160 Y Y S Mx E NA 14 R R 0 1.166666667 HB172 Y Y M Mx F/E NA 10 A R 0 0.833333333 HB99 Y Y M Ep E mut 7 DOD B 1 0.583333333 HB130 Y N S Mx NA mut 62 A R 0 5.166666667 HB98 Y Y S Ep M wt (FAP) 30 A M 0 2.5 HB136 N N S Mx F wt 34 A R 0 2.833333333 HB165 N N M Mx F/E mut 4 A R 0 0.333333333 HB1 Y Y M Ep E wt (FAP) 12 DOD R 1 1 HB93 N Y M Mx E mut 33 A LT 0 2.75 HB129 N N S Mx E wt (FAP) 54 DOD R 1 4.5 HB33 N Y M Ep CF wt(AX1N1) 3.5 DOD R 1 0.291666667 HB100 N N S Ep F mut 20 A B 0 1.666666667 HB184 Y Y M Ep E NA 14 DOD LT 1 1.166666667 HB157 Y N M Ep CF mut 5 R LT 0 0.416666667 HB80 Y Y S Ep CF mut 14 DOD R 1 1.166666667 HB86 N Y S Ep E mut 57 A R 0 4.75 HB8 N Y S Ep E mut 135 A R 0 11.25 HB147 N N S Mx F NA 12 A R 0 1 HB39 N Y S Mx NA mut 66 A R 0 5.5

TABLE C Gene Name AFP ALDH2 APCS AP0C4 AQP9 BUB1 C1S CYP2E1 DLG7 HC161 2.079447 −5.920384 −6.086912 −7.366206 −7.320175 4.176845 −6.502865 9.12672475 5.322878 HC162 4.056751 −3.64102 −4.586098 −5.663246 −4.233021 3.559124 −4.64283 −4.136919 5.950173 HC163 3.323238 −6.086663 −6.399079 −4.052853 −6.010302 4.772507 −6.776158 −8.515956 5.551408 HC164 3.075226 −6.146711 −7.241796 −3.371322 −5.446966 3.634476 −7.462807 −5.829384 3.98399 HC165 2.685177 −7.0470725 −6.294538 −7.242275 −6.94561 4.029514 −5.926596 −3.033642 5.723743 HC168 1.501031 −6.016314 −6.696324 −5.130347 −5.64774 3.305894 −6.883263 −4.411302 4.362859 HC169 2.880925 −6.024682 −6.87168 −4.19185 −6.058572 4.09117 −6.767215 −8.63753 4.614352 HC170 2.3753035 −6.6226955 −8.3702955 −5.4072375 −5.6954625 5.5639145 −8.0538815 −9.7948605 6.6275145 HC171 3.001804 −2.573977 −4.213123 −4.040859 −4.992701 3.583809 −5.226561 1.25382 3.874142 HC172 1.164528 −5.314302 −6.094852 −4.127298 −3.890072 3.991173 −6.240002 2.279678 5.651484 HC173 4.694127 −6.373823 −5.51865 −6.056863 −6.314031 4.30288 −4.863168 −8.649852 5.564261 HC176 4.066485 −5.552505 −5.444218 −5.551191 −5.815727 6.073568 −5.850428 −9.402043 6.051409 HC177 2.692613 −5.43842 −3.091896 −4.656336 −5.907612 3.452047 −6.412596 −10.50172 4.083836 HC178 −0.554213 −5.646708 −7.296414 −4.588115 −5.579087 3.125179 −6.556397 −6.591304 4.755443 HC179 1.910595 −4.139932 −8.136252 −6.036987 −2.847761 3.895205 −4.943672 −5.283326 5.054346 HC180 3.212685 −5.831134 −7.519348 −5.962761 −6.611712 1.5179 −6.130592 −9.203789 2.22658 HC181 6.030393 −4.04397 −2.03808 −0.956533 −2.850753 5.430957 −4.712002 −2.555649 5.031845 HC182 3.376941 −7.072651 −7.74873 −5.2003 −5.445893 6.665657 −7.899793 −10.089271 7.487442 HC183 3.149578 −4.684626 −7.045155 −3.800078 −7.042931 2.40337 −6.412624 −9.657513 3.396236 HC184 −0.093476 5.985909 −7.203484 5.482853 −6.208594 1.558788 −6.347367 −9.658434 2.407985 HC185 1.405595 −4.748444 −5.89589 −3.780913 −2.802368 4.37289 −5.800822 −5.410746 4.6459 HC186 1.666457 −5.52819 −7.953401 −3.287374 −3.805233 1.040678 −7.309734 −6.699831 2.197157 HC187 3.652111 −4.151991 −7.459358 −6.247812 −5.346647 4.211928 −6.33068 −8.629261 4.520672 HC188 0.355562 −5.261937 −7.83848 −4.759525 −4.839348 5.111208 −7.787661 −4.575966 5.635841 HC189 1.239891 −4.501697 −8.737075 −6.152778 −6.402122 5.0291015 −6.951675 −5.450079 4.419359 HC190 3.306642 −4.365515 −7.399538 −4.721411 −6.178224 3.016906 −4.970499 −5.850237 9.264351 Gene Name DUSP9 E2F5 GHR HPD IGSF1 NLE RPL10A HC161 3.702615 1.025512 −0.817005 −7.653863 14.149408 5.1985405 −5.81852 HC162 1.738977 1.432598 −0.231753 −6.700146 14.781699 1.231146 −5.9665735 HC163 4.00436 1.072797 −2.746621 −6.213082 8.2477055 2.203781 −5.49725 HC164 4.25604 2.567639 −3.606813 −6.079645 12.649441 1.946926 −5.171041 HC165 1.788757 1.157215 −1.197022 −7.969042 14.270796 2.620134 −6.219366 HC168 5.625335 2.2963 −1.169362 −7.52548 8.041574 2.337152 −5.42627 HC169 3.838008 1.60884 −2.921191 −6.51064 8.136143 2.099644 −5.731897 HC170 1.8626715 1.6955475 −3.9034625 −7.4271305 7.756398 2.6917235 −5.8132855 HC171 5.349357 2.074272 −1.437519 −5.297939 6.325863 3.057537 −3.95361 HC172 5.592005 1.291773 −0.040049 −6.989866 6.998259 3.186024 −3.946432 HC173 4.718896 1.367846 −2.3934 −7.781412 9.1259525 1.82226 −4.957916 HC176 2.248373 2.709599 −3.2392 −7.594156 7.5288985 1.817325 −5.042318 HC177 −0.297108 2.149313 −2.166834 −7.847734 5.8240705 1.530536 −5.640103 HC178 4.943904 1.038474 −1.620902 −5.659262 5.416822 1.855914 −4.954215 HC179 1.464274 1.372578 −0.386778 −6.31274 7.244471 1.887378 −5.218281 HC180 0.161194 −0.215954 −0.371454 −6.978048 5.185486 1.004282 −6.187635 HC181 4.322323 2.990459 2.18165 −0.651095 4.292234 4.670446 −2.978533 HC182 2.395117 2.329727 −4.420263 −7.357922 7.932783 2.869667 −5.574881 HC183 3.7002215 −0.85541 0.078707 −7.143723 11.999761 0.63414 −6.105039 HC184 2.266351 6.244093 0.670045 −6.27671 6.935964 1.564672 −6.568913 HC185 1.811225 2.225761 −1.246884 −7.344763 10.1413645 1.39443 −5.015711 HC186 −2.717975 1.183123 −2.657936 −7.680597 8.921477 1.289946 −6.631908 HC187 −0.066629 −2.0378 1.078709 −8.251018 7.478678 1.655093 −5.763416 HC188 1.839584 0.638515 −1.989428 −6.736329 12.8628775 2.27923 −4.743699 HC189 6.509026 −0.7698 −2.238756 −8.600128 11.305903 −0.437812 −7.061492 HC190 0.70722 4.181534 −0.773062 −4.881306 2.422048 −5.53509

TABLE D Table of normalized qPCR data (deltaCt values) of 88 HCCs analyzed by the Taqman method Gene name AFP ALDH2 AP0C4 APCS AQP9 BUB1 C1S CYP2E1 HC 001 2.212911 −6.2372335 −0.614689 −7.0721355 −6.047695 3.841505 −8.163492 −10.3093235 HC 003 3.865709 −6.230074 −0.95786 −7.52919 −6.7334475 0.147459 −8.7963405 −10.428074 HC 004 7.6758115 −2.186358 1.608247 −5.845683 −3.759528 4.221132 −5.8997645 −7.1147515 HC 006 7.9469815 −5.4231035 −0.9614255 −7.3704745 −7.006052 0.5252045 −8.162856 −10.1334265 HC 007 −5.311541 −4.0446765 3.550537 −5.1967915 −6.747103 0.299039 −4.062593 −11.024027 HC 008 −2.0890815 −3.9297005 0.6776965 −6.567126 −3.1082155 1.214781 −7.2991535 −7.791007S HC 009 7.0483095 −3.0017225 9.6721075 0.017488 −3.7536735 −2.980029 −4.830331 −0.5825245 HC 010 −2.3869635 −0.95212 0 1.0272875 −1.3400495 1.864677 −2.639902 −3.604805 HC 011 −0.6488335 −5.958108 −1.076151 −7.7638255 −6.122144 2.362454 −8.319293 −9.575619 HC 012 6.538312 −4.6271565 1.221393 −6.942673 −4.1878425 3.293346 −6.850023 −7.284587 HC 014 2.987769 −5.194577 −1.3542145 −6.5396565 −6.8623455 1.363697 −6.8939375 −10.7465595 HC 015 −6.14089 −4.5178635 5.156026 −3.380102 −2.373344 −0.8830545 −7.1343975 −4.9390935 HC 017 −7.1950405 −2.6522585 2.395651 −4.5167035 −2.8711295 −1.0884485 −6.035123 −6.037085 HC 018 6.856588 −1.840894 3.84764 −4.916924 −3.6093495 0.063545 −4.263272 −5.811062 HC 020 0.65281 −6.287083 −3.2094885 −8.2117635 −7.354605 1.4635025 −8.471663 −10.2536915 HC 021 4.3070475 −2.175112 6.2591235 −5.9159775 −1.1452535 −0.0802935 −5.7190985 −1.2878015 HC 022 4.418018 −5.331214 −0.5455545 −6.6835035 −5.7992305 2.173361 −7.2514145 −8.0876755 HC23 5.538438 −5.853486 −0.5708905 −6.9009145 −6.651868 2.5475915 −8.2212235 −9.047509 HC 025 3.90298 −6.162477 −1.834891 −8.798759 −8.758959 2.5679685 −8.5606875 −10.814935 HC 026 5.69175 −5.0135775 −0.2581675 −7.2072275 −3.8645965 −0.545363 −7.2351705 −0.671071 HC 027 0.626755 −5.6309605 −1.53158 −7.2809855 −5.4736555 0.8889165 −8.172076 −8.6350095 HC 028 0 −1.913778 6.0251725 −1.0475505 −0.9613895 5.7426525 −4.910584 −3.6858305 HC 030 −6.4370325 −3.8476295 −0.2797975 −7.1142435 −5.0250435 0.190936 −7.5279395 −7.5682115 HC 032 −0.0037145 −6.802666 −2.574347 −7.500133 −7.530391 5.1317805 −7.854502 −9.4408715 HC 034 6.6945705 −5.11617 −0.5860455 −7.134934 −6.9427395 1.2674215 −7.719763 −8.545814 HC 037 1.3519745 −5.808058 0.0768065 −6.755895 −6.3416265 2.4955985 −6.921051 −10.1686795 HC 038 −4.053435 −4.596143 0.129322 −5.045701 −6.0302545 −0.321483 −6.101331 −8.1123675 HC 041 2.7156435 −6.3503265 −2.281983 −5.612517 −7.8444565 0.587016 −6.88808 −9.5090495 HC 042 5.216493 −4.4086495 0.627239 −4.1054755 −6.063786 2.224818 −6.3060565 −9.1411555 HC 043 1.7983435 −5.457548 0.7055185 −7.607914 −4.7175855 2.8634735 −7.9862115 −8.760714 HC 052 −10.3337105 −2.1920375 8.124407 −5.9818015 0.4848805 1.2986035 −5.6337865 −1.7693015 HC 058 −1.891958 −2.1172735 11.8524 4.1106695 2.817265 −1.9395175 −3.691331 4.3317445 HC 060 −7.624821 −3.6860195 0.545509 −8.100997 −6.8503395 0.576028 −8.167253 −9.1875325 HC 064 −5.0266755 −4.992107 −0.7860345 −7.4148835 −7.0526325 1.367463 −7.1364365 −9.682147 HC 066 −3.156328 −3.8408415 0.6773785 −8.2106815 −6.2767975 1.1272665 −8.026875 −8.601088 HC101 6.873135 −4.339036 0.5787185 −6.288568 −4.6233735 −0.081457 −7.321092 −5.806032 HC102 4.119697 −2.476355 5.453696 2.3952165 −0.0196725 0.5553155 −5.939374 2.8566735 HC103 −1.6193685 −3.889904 0.54698 −6.014572 −7.151639 2.086008 −5.965432 −8.266311 HC104 −5.5094265 −4.936239 0.5059805 −5.624234 −0.501258 1.311194 −6.716137 −9.0888685 HC105 −2.3444245 −4.239726 3.577778 −7.703333 −4.2748785 −0.945674 −7.774455 −5.698899 HC106 3.42054 −6.1642895 0.7836775 −7.8462545 −5.85931 4.8909655 −8.060072 −9.9949555 HC107 4.136209 −6.7443095 −4.4534435 −9.2080655 −8.8878655 1.7415115 −9.2061165 −9.3234825 HC108 4.500336 −3.6076385 2.478085 −7.275462 −4.4353395 0.3807995 −7.1031155 −3.889942 HC109 4.833024 −5.8617665 −0.729565 −6.222909 −6.4504115 2.2918285 −7.406001 −8.7101925 HC110 3.5240185 −3.6707715 0.256479 −5.043319 −4.5999895 1.449943 −6.9163195 −7.145766 HC111 1.883473 −3.8304065 1.130067 −5.976754 −4.1657805 −0.621548 −6.278164 −4.46942 HC112 2.8803905 −4.8726745 0.7777655 −6.764675 −5.2735435 −0.3135015 −7.455794 −2.5741475 HC113 −1.208649 −4.407016 2.366969 −5.197177 −2.681192 3.4825665 −6.338901 −6.443846 HC114 5.4433695 −4.7113965 0.833543 −6.723142 −4.445291 1.7431855 −7.866014 −7.3429245 HC119 −1.0580855 −6.159706 −1.894453 −9.375177 −7.6266135 0.797564 −9.1461175 −7.095824 HC120 4.0065425 −4.257398 3.5241745 −5.6838965 −6.8239115 0.0740105 −8.5708615 −7.6044515 HC121 4.254961 −4.556431 2.167313 −6.2688205 −4.38702 2.4486685 −8.118416 −7.765037 HC122 2.3763095 −6.2844515 −1.279577 −6.9942545 −6.8198535 6.0183915 −7.7653135 −9.450349 HC123 −0.821555 −4.220769 0.68167 −5.778659 −6.410177 1.190323 −5.383781 −8.528543 HC124 −3.9525335 −4.027289 0.0499065 −5.391271 −4.463488 1.592563 −5.151686 −9.520436 HC125 4.806564 −4.5451465 −2.6326775 −6.5321595 −8.370224 −1.1627945 −8.4244055 −9.426232 HC126 5.899437 −5.02839 −0.407895 −5.2838365 −3.6163545 2.6943025 −7.1365955 −5.226091 HC127 0.0390765 −2.41699 −0.8680995 −4.846116 −1.8613935 2.048769 −6.3641695 −6.1813065 HC128 −5.8636305 −5.085525 0.626498 −5.087517 −4.3184915 1.3297375 −6.828468 −7.4344035 HC129 3.430757 −4.6298475 1.863955 −4.8448705 −2.870839 2.3688215 −7.302922 −2.692798 HC131 1.491189 −5.425994 −2.4702 −8.6617295 −7.4772145 0.727709 −7.525072 −8.98645 HC132 −5.4265205 −3.105643 6.9974515 3.2748865 −3.9244375 −0.2895395 −4.390082 −7.0455735 HC133 5.1621395 −4.2462915 −0.63156 −7.145861 −6.05182 4.9277675 −7.3188145 −8.1908895 HC134 −2.8738695 −4.061101 0.1134065 −7.5103485 −5.550642 −1.7425995 −8.4609335 −7.859701 HC135 0.909107 −2.7442165 0.7630605 −0.959726 −4.0595615 1.2018365 −4.667223 −4.30592 HC136 0.4105125 −6.0408575 −0.7390785 −7.150737 −5.996196 4.288554 −8.243333 −9.042865 HC137 −4.378388 −3.2913795 3.209294 −4.421328 −0.5225755 4.2185175 −5.647363 −5.532515 HC138 2.4762965 −4.8248625 1.154563 −4.883388 −3.440722 3.408251 −6.459976 −7.2458685 HC139 2.7547595 −2.9782295 3.0252085 −5.3858735 −5.0157665 0.9503045 −6.0281485 −1.1920485 HC140 6.3489955 −4.644452 −1.006979 −2.1507335 −5.3387635 4.075603 −6.7373815 6.646618 HC141 2.4010865 −4.8883675 0.787009 −4.7365085 −4.1224775 4.2728925 −6.8664705 −2.6765195 HC142 4.5984525 −3.7946485 2.8271835 −4.9243665 −3.1411815 4.0713025 −6.3482925 2.654871 HC143 −4.0727165 −2.59764 1.855993 −4.8795135 −2.222047 1.6908025 −4.948264 −3.1057735 HC144 4.7344185 −4.3542505 −1.002913 −0.432856 −5.16696 2.510931 −5.3365195 −4.456082 HC145 8.5175565 −3.375805 0.8672075 −5.0765195 −4.091142 3.9700095 −6.960951 0.8009 HC146 5.741507 −3.5738745 1.2439275 −5.1950135 −3.4305425 2.9843625 −5.666896 −0.913546 HC147 6.0474775 −3.0470955 0.2246755 −5.6213855 −5.257189 2.7534355 −5.349428 −6.933909 HC148 −1.306432 −4.0108565 0.267747 −6.3544915 −3.1846315 1.1995135 −6.2066555 −4.1428355 HC149 −3.9190605 −3.3456535 2.735403 −1.9099995 −1.1810265 2.704253 −5.707004 −5.9300895 HC150 6.1556695 −2.9923905 −1.9485835 −5.821769 −6.3127705 2.452404 −4.984573 −7.3184395 HC151 5.5488065 −4.234966 1.372415 −5.8812085 −4.0297925 3.4239945 −7.2861515 −2.304461 HC152 4.917902 −3.97386 −4.005999 −6.5072455 −7.124415 2.5576145 −5.752235 −9.98327 HC153 5.6708455 −5.004032 −3.204075 −3.8195495 −6.2020215 1.9670395 −5.979251 −7.7421455 HC154 6.699114 −2.0392575 9.6136985 0.885791 −0.68511 1.755108 −0.7395055 2.544628 HC155 6.238831 −3.802053 2.0022335 −6.3105565 −2.974712 4.2276825 −7.058571 −4.1514335 HC156 −1.582839 −3.5688085 0.917505 −3.9333845 −4.163765 1.0763025 −4.6064345 −8.4802835 HC157 3.657864 −4.2315665 2.513598 −7.2096625 −4.573216 −0.284071 −5.856564 −7.9837885 HC159 3.4650565 −2.6801805 2.2596385 −4.0834345 −4.42904 3.44645 −5.923485 −7.778452 Gene name AFP ALDH2 AP0C4 APCS AQP9 BUB1 C1S CYP2E1 HC 001 5.30317 11.616567 −0.05328 −2.655512 −9.449416 6.46034 1.159417 −6.6225235 HC 003 2.057513 8.8462855 1.909804 −2.069524 −8.549803 7.249974 1.5801355 −6.0562915 HC 004 4.4226465 9.4268185 1.7432195 2.0012965 −9.415253 0 3.1459935 −4.4121905 HC 006 1.6282005 10.22051 −0.024339 −1.887805 −8.5958965 7.1580385 −0.6940375 −6.8637555 HC 007 1.169221 6.6521625 0.2833465 1.7428205 −6.183977 3.192514 0.3919565 −7.1381125 HC 008 2.80866 9.6946695 0.0193165 −2.342442 −5.329776 2.806768 1.579419 −6.2574845 HC 009 −1.3733475 9.5262655 −0.711082 2.3242195 0.011478 4.026769 0.80375 −6.3016635 HC 010 0 0 1.344368 0.4900285 −2.932809 0 0 −9.1966395 HC 011 2.8432205 0 0.736822 −4.757848 −9.029214 7.6390015 1.9328755 −7.379063 HC 012 4.7199665 0 2.4002515 −2.2402875 −9.656029 7.466951 1.64183 −5.178571 HC 014 3.3543285 7.7629895 1.5332515 −1.09511 −9.5837645 8.5836025 1.47219 −5.831244 HC 015 0.1414205 4.4342765 −1.399564 −0.2426 −4.473096 −0.0722075 0.321593 −6.8777395 HC 017 −0.666284 3.163581 −1.206766 2.353691 −0.6808655 6.0490105 0.386649 −7.068098 HC 018 1.512286 8.7756845 2.426129 2.9035 −5.7101575 2.4248235 1.3815525 −5.9464565 HC 020 2.1165725 9.6208445 1.1944835 −4.5756335 −10.6864405 0 1.118745 −7.542193 HC 021 0.322455 7.8162765 0.0686475 −0.71981 −4.0108195 2.954814 1.618369 −6.309556 HC 022 3.3904095 10.82729 10.7133385 −2.416651 −9.8859985 5.6986975 1.9449755 −7.194012 HC23 3.848364 0 1.4330655 −3.7226655 −9.583194 7.200325 1.823275 −5.9526365 HC 025 3.34202 7.1111525 −0.049846 −1.9012935 −9.1845675 0 1.770127 −7.4507165 HC 026 0.9710395 8.5287915 1.1845665 −1.964045 −7.6403735 5.4960635 1.851733 −5.9670715 HC 027 2.3158215 10.241011 0.4045835 −2.623084 −9.597772 5.588995 1.851285 −7.6623025 HC 028 0 0 4.334386 1.9788575 −3.3142495 0 2.4559905 −5.521873 HC 030 0.189092 9.0027 −1.0623035 −2.635437 −7.537 2.651022 1.2674865 −7.5046195 HC 032 5.7080765 9.73163 0.054818 −2.0027475 −9.0015185 0 1.208576 −8.8437815 HC 034 2.339621 9.9728495 1.4281575 −1.563203 −8.3685675 10.112616 1.934745 −6.594006 HC 037 2.6534895 0 1.2212655 −2.9415775 −10.367265 7.5570255 1.9881245 −6.901637 HC 038 1.4386515 5.2298755 0.037887 −0.2025015 −7.547286 0.680358 2.1250395 −5.1574215 HC 041 1.840185 8.727439 −0.466649 −1.428749 −8.0015745 7.243446 0.15624 −7.7043325 HC 042 3.2531575 0 0.3673235 1.2545195 −8.2669835 2.899766 0.9401045 −5.577659 HC 043 4.2390495 10.525647 0.894345 −3.2916395 −8.997825 5.5544715 1.8422595 −5.480403 HC 052 2.599359 3.8059605 −0.4419525 1.843696 −2.481945 −2.254168 1.9474305 −5.6154705 HC 058 −0.1957495 3.656912 −0.804087 3.7242975 −1.8257985 −1.3471695 1.209522 −6.0601515 HC 060 2.2644225 6.618755 0.432422 1.4079225 −8.4643875 0.7884805 1.9133155 −5.7041285 HC 064 2.386875 7.3184655 0.2876185 −0.349645 −8.6027575 3.3382005 1.817699 −6.4617635 HC 066 2.7680135 11.5673955 0.968982 1.2501855 −8.5231325 9.185554 1.962008 −5.415169 HC101 1.3084655 8.828389 1.871516 −0.1466275 −5.7252795 4.1394545 1.4546305 −6.144011 HC102 2.1385165 8.6628475 −0.830934 −0.947389 −0.568809 2.708733 1.1534675 −5.283399 HC103 2.957914 12.521336 1.8003215 −0.636723 −6.717282 9.802921 2.594702 −4.423835 HC104 1.821739 5.396553 2.305498 −1.6860905 −8.46781 −0.1438735 1.610158 −6.21159 HC105 0.814912 5.4214725 −2.0730715 −0.682142 −2.288109 1.422332 0.471391 −6.315756 HC106 6.2678815 11.174152 2.208171 −5.342392 −9.4440475 7.401009 1.968983 −5.769397 HC107 1.357756 6.6136855 −2.78876 −2.935929 −10.460972 0 0.000835 −8.6686655 HC105 2.2445545 8.0946735 −0.0923905 −1.6363755 −2.9674235 7.967992 0.932052 −5.818028 HC109 3.222524 10.4709205 1.9924345 −2.9233285 −7.8859205 10.0122565 2.6102395 −5.541229 HC110 2.333076 11.616244 2.512512 −1.0803015 −8.1908235 8.1469415 2.3529485 −5.245476 HC111 0.769283 9.137462 −1.045678 −1.1576425 −7.245347 1.86965 1.012752 −5.568205 HC112 0.9196845 10.105965 −0.0373705 −2.5391085 −7.714358 3.4428695 1.119237 −6.1905075 HC113 4.5602875 7.8299455 2.82243 −2.16232 −6.685692 2.045068 2.156348 −5.8884625 HC114 3.1500875 11.804112 0.0450475 −2.5053965 −6.835254 5.1813245 1.3170345 −5.795905 HC119 1.712686 9.106547 0.0248045 −3.7649595 −9.220498 5.39017 0.400823 −7.954231 HC120 1.9563135 5.8119685 −1.229768 −3.196589 −8.5127155 9.404196 1.1096815 −6.4517175 HC121 2.852561 9.706684 0.910943 −2.2774645 −7.480725 5.980435 1.758163 −6.4042545 HC122 7.228946 9.9054825 3.5033365 −2.400201 −8.7301975 8.6480295 2.2430545 −5.199782 HC123 2.929576 11.584458 0.646839 1.810364 −4.7774665 5.1400615 1.5951645 −4.7323885 HC124 2.03781 8.81055 −0.574165 −2.2369305 −7.832169 1.4450915 0.1499775 −6.691521 HC125 −0.3286545 9.3740615 0.028878 −0.697866 −5.7813 10.2234745 0.405397 −7.1196575 HC128 3.944339 8.7174575 3.271927 −1.824385 −1.865621 7.659377 2.033278 −5.389272 HC127 2.96212 8.672372 2.162602 −0.129431 −3.4481965 3.1503205 2.205965 −4.3385115 HC128 2.6299155 8.499355 4.393094 −1.9716885 −5.7052855 2.72995 1.949352 −6.6181545 HC129 3.6405185 7.0627455 0.470421 −2.332961 −5.502918 5.692623 1.683808 −4.8697295 HC131 1.461713 8.415907 −0.154573 −4.009655 −8.960383 7.5832005 1.5313675 −6.775249 HC132 1.5572645 3.3843145 −1.9018925 −1.7710325 −2.3653865 1.947055 −0.2035885 −6.7796075 HC133 5.5447335 8.022457 2.6341825 −2.2298335 −6.1281315 0 1.4173895 −5.762015 HC134 −0.8148735 4.96739 −3.1030595 −1.3138565 −7.231144 0.3848995 −0.794433 −7.7140665 HC135 2.250305 5.794605 −0.986165 0.6955465 −6.7262275 4.394354 0.9780515 −6.689595 HC136 5.5267715 10.9307725 2.4040865 −4.013948 −8.223611 7.4962365 2.426321 −5.5069335 HC137 5.2105355 4.767228 5.62451 −1.6355645 −5.8875425 1.0556075 3.7311615 −5.2271275 HC 138 5.028429 5.576937 4.1601375 −1.738341 −6.019837 7.169314 4.19882 −4.2322595 HC 139 2.940447 4.3133685 0.685194 1.632571 −4.6240035 3.333358 1.7913325 −6.6866335 HC 140 5.1767035 10.874029 2.488357 −3.1717235 −7.5439415 9.276635 5.0732625 −4.266519 HC 141 6.1148255 7.979559 2.66802 −1.687093 −7.2596615 #DIV/0! 3.5973445 −4.952551 HC 142 5.8031125 8.2104255 2.0983905 −1.5934495 −5.8074755 9.442329 3.4164995 4.6520795 HC 143 3.470906 3.981805 1.474377 0.695168 −2.049901 3.754627 3.058019 −4.7443975 HC 144 3.844786 10.7187705 3.540563 −1.6857605 −6.869217 11.9441575 4.417722 −4.817306 HC 145 5.482263 9.313039 2.112409 −1.525041 −6.669204 10.0458615 3.0082705 −5.7677005 HC 148 5.1824885 7.611916 2.8802325 −1.791636 −6.9831945 5.450716 3.884913 −4.427413 HC 147 4.5366875 9.358894 3.2373475 −2.0156545 −6.053345 8.7065355 3.732017 −4.317148 HC 148 2.490156 5.4985645 8.523611 −0.773246 −3.7206575 5.663583 3.295068 −6.0532135 HC 149 3.4454215 6.8563245 2.4724295 −0.9357605 −7.337568 −0.063395 4.267075 −5.7767065 HC 150 3.585447 7.980274 3.118546 0.5916635 −5.762837 9.1651835 2.811495 −5.7495535 HC 151 4.613043 8.9062765 2.2090065 −2.8000785 −7.251033 9.44137 3.5959505 −4.6972005 HC 152 4.17552 10.736246 4.56538 −1.578246 −8.106859 12.118351 2.6658355 −6.944767 HC 153 3.133394 7.298329 3.85894 −0.616143 −7.947464 11.674272 2.670245 −5.0796695 HC 154 3.2541115 3.139705 −0.3936805 −1.070278 −4.611328 1.5925535 2.2396475 −6.2090535 HC 155 5.7341595 6.4585135 2.4375015 −0.254649 −7.297162 10.0981895 3.3878795 −5.37231 HC 156 2.1302465 4.4056075 1.070339 0.42868 −6.890963 2.0124875 2.225275 −7.037827 HC 157 1.3778545 2.0950385 −0.56173 −0.8411435 −8.474893 7.2842685 1.6720135 −6.6310375 HC 159 5.727853 8.8523415 2.7886015 −1.0442865 −7.268645 8.8204775 2.861685 −5.4777465

TABLE G Date of follow- Date of 1st secondary date of tumor HC 000 tumor surgery or (PH) or Date of Date of up recur- recurrence or OLT after secondary grade identification transplantation (OLT) last visit death (years) rence metastasis hepatectomy OLT Edmondson HC 001 12/12/1996 PH 07/01/1997 0.07 N 3 HC 003 21/02/1997 PH 20/06/2000 3.33 Y  4/11/1998 N 2 HC 004 28/02/1997 PH 20/08/2008 11.48 N 2 HC 006 07/10/1996 PH 06/01/1998 1.25 N 28/11/1997 N 2 HC 007 02/07/1996 PH 31/12/1997 1.50 Y  4/11/1997 N 2-3 HC 008 05/06/1996 PH 24/01/2005 8.48 N 3 HC 009 28/08/1996 PH 05/09/1996 0.02 N 3-4 HC 010 10/10/1996 PH 20/09/1997 0.95 N 4 HC 011 10/10/1996 OLT 14/12/2008 12.20 N 2 HC 012 24/10/1995 OLT 14/11/1995 0.05 N 2 HC 014 10/06/1995 OLT 27/07/1995 1.00 N 3-4 HC 015 21/07/1995 PH 10/10/1996 1.22 Y 10/10/1996 N 3 HC 017 05/05/1997 PH 16/04/2008 10.96 N 2 HC 018 07/05/1997 PH 28/09/1997 0.39 NA 3 HC 020 13/05/1993 OLT 20/10/2008 15.40 N 2 HC 021 15/01/1992 PH 28/09/1992 0.70 Y 15/06/1992 N NA HC 022 15/03/1997 OLT 02/09/2008 11.50 N 2 HC 023 20/07/1995 PH 20/06/2007 11.93 N 2 HC 025 05/10/1992 PH 13/08/2008 15.87 N 2 HC 026 04/06/1993 OLT 18/04/1994 0.83 NA 2 HC 027 20/01/1993 OLT 15/02/1993 0.10 N 2 HC 028 16/02/1996 OLT 13/03/1996 0.10 N 3 HC 030 10/04/1996 PH 07/09/2008 12.40 Y 15/10/1996 Y 17/12/1993 3 HC 032 17/02/1993 PH 17/10/1993 0.66 N 2 HC 034 10/03/1993 PH 05/11/2008 15.70 Y 15/11/1995 Y 20/06/1996 2 HC 037 08/06/1997 OLT 13/08/1997 0.20 N 3 HC 038 16/07/1997 PH 28/08/1998 1.12 Y  1/01/1998 N NA HC 041 24/11/1997 PH 01/05/2005 7.44 Y 29/06/1999 Y  9/3/2000 2 2nd recurrence 15/1/2005 HC 042 05/11/1997 PH 03/06/2008 10.58 N 3 HC 043 19/11/1997 OLT 22/10/2008 10.90 N 3 HC 052 17/02/1999 PH 18/05/1999 PDV 0.25 N 3 HC 058 14/10/1999 PH 30/01/2008 8.30 N 2 HC 060 15/05/1925 PH NA NA HC 064 10/04/2000 PH 09/07/2005 5.25 Y 15/10/2001 N 3 HC 066 15/09/1999 PH 18/08/2008 8.93 N 2-3 HC 101 03/05/2006 0LT 27/10/2008 2.50 N 2-3 HC 102 12/07/2006 PH 18/08/2006 0.10 N 4 HC 103 16/08/2006 PH 11/06/2008 1.82 Y 15/1/2007  N 2-3 HC 104 20/09/2006 PH 05/11/2008 2.10 N 2-3 HC 105 11/12/2006 PH 04/07/2007 0.56 Y 15/04/2007 N 3 HC 106 22/01/2007 OLT 16/01/2009 2.00 Y 3 HC 107 25/01/2007 PH 23/10/2008 1.75 N 2 HC 108 12/02/2007 PH 24/09/2008 1.62 N 3 HC 109 19/02/2007 OLT 26/05/2008 1.30 N 2-3 HC 110  6/02/2007 OLT 04/02/2009 1.95 N 2-3 HC 111 07/03/2007 OLT 03/10/2007 0.70 N 2-3 HC 112 19/03/2007 PH 08/09/2008 1.48 N 2-3 HC 113 23/03/2007 OLT 15/03/2008 1.00 N 2-3 HC 114 03/04/2007 PH 11/09/2007 0.44 N 2 HC 115 01/08/2007 PH 29/04/2008 0.75 N 1 HC 116 09/08/2008 PH 18/04/2008 0.69 N 3 HC 117 25/10/2007 OLT 23/12/2008 1.20 N 2-3 HC 118 25/10/2007 PH 28/09/2008 0.93 N 1 HC 119 03/12/2007 OLT 08/01/2009 1.20 N 2-3 HC 120 18/12/2007 PH 14/10/2008 0.82 N Y 12/05/2008 2-3 HC 121 02/01/2008 PH 08/08/2008 0.60 N 3 HC 122 16/01/2008 PH 17/10/2008 0.75 Y 10/10/2008 N 2 HC 123 11/02/2008 OLT 01/12/2008 0.80 N 3 HC 124 20/02/2008 PH 26/08/2008 0.52 N 3 HC 125 22/02/2008 OLT 08/01/2009 0.90 N 3 HC 126 12/03/2008 PH 14/08/2008 0.42 Y 6/8/2008 N 1-2 HC 127 19/03/2008 PH 20/06/2008 0.25 Y 4/6/2008 N 2-3 HC 128 20/03/2008 PH 29/08/2008 0.44 N 2 HC 129 01/04/2008 0LT 31/05/2008 0.15 N 3 HC 130 07/04/2008 PH 27/05/2008 0.14 N 3 HC 131 10/04/2008 PH 15/07/2008 0.26 N 2-3 HC 137 19/07/2002 PH 31/03/2008 . 5.67 N . NA HC 138 25/04/2003 PH 03/12/2008 . 5.58 Y 03/10/2003 NA HC 139 15/05/2002 PH 09/05/2008 . 6.00 N . NA HC 140 03/06/2004 PH  5/08/2008 . 4.17 Y 30/06/2005 NA HC 141 06/02/2004 PH 12/03/2009 . 5.08 Y Dec. 2005 NA HC 142 14/05/2002 PH 21/06/2006 21/06/2006 4.08 Y 24/03/2006 NA HC 143 04/03/2002 PH 26/01/2007 . 2.83 Y 2005 NA HC 144 27/06/2002 PH 17/06/2008 . 6.00 Y 16/03/2004 NA HC 145 14/11/2002 PH 30/07/2008 . 5.58 Y 09/06/2005 NA HC 146 30/07/2004 PH 11/12/2008 . 4.33 Y June 2005 NA HC 147 23/11/2004 PH 22/09/2008 . 3.83 Y 12/06/2008 NA HC 148 12/09/2003 PH 15/10/2006 . 3.08 N NA HC 149 26/08/2003 PH 16/01/2007 16/01/2007 3.42 N NA HC 150 31/01/2003 PH 23/06/2008 . 5.42 N NA HC 151 10/12/2004 PH 15/03/2007 . 2.25 N NA HC 152 14/05/2003 PH 17/01/2007 17/01/2007 3.67 Y mars-09 NA HC 153 25/02/2003 PH 24/12/2007 24/12/2007 4.83 Y 06/05/2005 NA HC 154 06/09/2004 PH 23/11/2006 2.21 Y 01/01/2005 N 2-3 HC 155 18/10/2004 PH 09/12/2008 4.10 Y 18/10/2004 Y 31/05/2005 2 HC 156 03/02/2005 PH 28/05/2007 2.31 Y 15/06/2006 3 HC 157 24/02/2003 PH 26/10/2006 3.59 Y 15/08/2004 2 HC 159 16/10/2002 PH 18/03/2005 2.42 Y 03/05/2004 2 HC 161 20/08/2003 PH 06/02/2008 4.47 Y 2 HC 162 30/10/2003 PH 25/04/2007 3.49 N 3 HC 163 20/09/2004 PH 07/12/2006 2.21 Y 01/09/2006 N 3 HC 164 05/09/2002 PH 21/03/2007 4.54 N 1 HC 165 08/08/2003 PH 29/05/2008 4.72 N 2 HC 168 10/02/2003 PH 04/02/2009 6.00 Y 15/07/2004 Y 18/02/2008 2 HC 169 10/06/2002 PH 22/03/2005 22/03/2005 2.78 Y 15/03/2003 N 2 HC 170 14/03/2002 PH 28/06/2007 5.29 N 1 HC 171 25/03/2004 PH 17/10/2008 4.57 Y 15/11/2004 N 4 HC 172 10/01/2005 PH 25/11/2008 3.90 Y 25/11/2005 N 3 HC 173 18/12/2003 PH 03/03/2008 4.21 N 1 HC 176 13/03/2002 PH 05/10/2006 4.57 N 2 HC 177 29/10/2003 PH mars-09 5.42 Y 01/2009 2 HC 178 19/03/2003 PH 19/09/2005 2.50 N 2 HC 179 27/10/2000 PH 06/12/2005 5.17 Y 10/2002 2-3 HC 180 9/4/2002 PH 03/11/2005 03/11/2005 3.58 Y 05/2005 3 HC 181 27/05/2002 PH mars-09 6.83 Y 04/2008 2 HC 182 30/03/2004 PH October 3.50 N 1 2007 HC 183 21/07/2003 PH 02/09/2007 02/09/2007 4.08 Y July 2007 3 HC 184 18/01/2002 PH 08/02/2004 08/02/2004 2.08 Y April 2002 2 HC 185 19/11/2002 PH 03/03/2005 2.25 N 3 HC 186 31/08/2004 PH 06/11/2006 06/11/2006 2.17 N 3 HC 187 7/06/2001 PH févr-09 7.67 Y March 2003 1 HC 188 29/07/2004 PH avr-09 4.67 Y July 2004 2 HC 189 30/04/2002 PH 13/08/2005 13/08/2005 3.25 Y January 2 2005 HC 190 29/07/2003 PH mars-09 5.58 N 3 number max of Macro- mitosis Ndules Nrmal Score Score Tumor vascular vascular per 10 of liver Cir- META- META- HC 000 tumor differenti- tumor invasion invasion fields × multiple regen- A0F0 or rhosis VIR VIR identification ation (OMS) size (mm) macro micro 40 Ndules eration A0F1 AXF4 Activity Fibrosis HC 001 moderately 120 N N NA N N Y NA 4 differentiated HC 003 well 60 N N NA N N Y NA 4 differentiated HC 004 well 100 N N NA N Y N 0 1 differentiated HC 006 well 90 N Y NA N Y N 0 1 differentiated HC 007 well 100 Y Y NA Y N N 2 3 differentiated HC 008 moderately 30 N N NA N N Y N 4 differentiated HC 009 Moderately 100 Y Y NA Y N N 1 3 poorly HC 010 moderately- 75 N N NA N N Y NA 4 poorly HC 011 well 15 N N NA Y N Y NA 4 differentiated HC 012 well 60 N N NA Y N Y NA 4 differentiated HC 014 Moderate 80 Y Y NA Y N Y NA 4 poor HC 015 moderately 60 Y Y NA Y N N 3 3 differentiated HC 017 well 100 N N NA N N N NA 3 differentiated HC 018 moderately 140 Y Y NA N N Y 2 4 differentiated HC 020 well 40 NA NA NA Y N Y NA 4 differentiated HC 021 NA 100 NA NA NA Y N Y NA 4 HC 022 well 45 N N NA Y N Y NA 4 differentiated HC 023 well 50 N N NA N Y N NA 0 differentiated HC 025 well 140 N N NA N Y N 0 0 differentiated HC 026 well 30 Y Y NA Y N Y NA 4 differentiated HC 027 well 15 N N NA Y Y N Y NA 4 differentiated HC 028 moderately 120 N Y NA Y Y N 0 0 differentiated HC 030 moderately 16 NA NA NA N N Y NA 4 differentiated HC 032 well 60 N NA NA Y N Y NA 4 differentiated HC 034 well 140 N N NA Y Y N NA 0 differentiated HC 037 moderately 35 Y Y NA Y Y N Y NA 4 differentiated HC 038 moderately 50 N N NA Y N Y NA 4 differentiated HC 041 well 30 N N NA N N Y NA 4 differentiated HC 042 moderately 130 prob- Y NA N N N 2 1 differentiated able HC 043 moderately 15 N N NA Y N Y N 4 differentiated HC 052 moderately 110 N Y NA Y N Y N 4 differentiated HC 058 moderately 100 N N NA N N N 2 3 differentiated HC 060 well 55 N N NA differentiated HC 064 moderately 40 N N NA N N N 2 2 differentiated HC 066 well 75 N N NA Y N Y NA 4 moderately HC 101 well 35 Y Y 18 Y Y N Y 2 4 moderately HC 102 Peu 200 Y Y 7 N N N N 1 1 différencié HC 103 well 55 N Y 8 N Y N Y 3 4 moderately HC 104 well 160 prob- Y 10 Y N Y N 0 1 moderately able HC 105 moderately 40 Y Y 20 Y Y N Y 2 4 differentiated HC 106 moderately 80 Y Y 32 Y N N Y 1 4 differentiated HC 107 well 60 N N 1 N N Y N 0 0-1 differentiated HC 108 moderately 26 N Y 18 N N N N 1 1 differentiated HC 109 well 30 N N <1 Y Y N Y 2 4 moderately HC 110 well 30 N Y 1á5 Y Y N Y 1 4 moderately HC 111 well 40 Y Y 45 Y Y N Y 1 4 moderately HC 112 well 18 N N 0 N N N N 2 2 moderately HC 113 well 50 Y Y 25 Y Y N Y 1 4 moderately HC 114 well 36 N N <1 N N N N 2 3 differentiated HC 115 well 90 N N 0 N N N N 2 1 differentiated HC 116 moderately 140 N N 12 N N N N 2 3 differentiated HC 117 well 28 N N 4 Y Y N Y 2 4 moderately HC 118 well 40 N N <1 N N Y N 0 1 differentiated HC 119 well 26 N Y 15 Y Y N Y 2 4 moderately HC 120 well 20 N Y 3 Y N N Y 1 4 moderately HC 121 moderately 150 prob- Y 8á30 Y Y N Y 2 4 differentiated able HC 122 well 20 Y Y 8 Y ? N Y 1 4 differentiated HC 123 moderately 43 prob- prob- 4 Y N N Y 2 4 differentiated able able HC 124 moderately 62 N N 4 N N N N 1 1 differentiated HC 125 moderately 33 N Y 2 Y N N Y 2 4 differentiated HC 126 well 130 Y Y 2 Y N Y N 0 1 differentiated HC 127 well 115 Y Y >100 N N N N 1 1 moderately HC 128 well 110 N Y 5 N N N N 2 2 moderately HC 129 moderately 30 N Y 40 Y N N N 2 3 differentiated HC 130 moderately 38 N prob- 12 N N N N 1 2 differentiated able HC 131 well 120 N Y 20á25 N N Y N 0 1 moderately HC 137 moderately 10 NA NA NA Y . . . differentiated HC 138 well 5.5 NA NA NA Y N . . differentiated HC 139 moderately 16 NA NA NA Y . . . differentiated HC 140 well 15 NA NA NA N N 0 1 differentiated HC 141 well 3.5 NA NA NA N N . . differentiated HC 142 well 8 NA NA NA Y . . . differentiated HC 143 well 3 NA NA NA N Y 1 4 differentiated HC 144 well 15 NA NA NA Y . . . differentiated HC 145 well 6 NA NA NA N . 0 3 differentiated HC 146 well 7.5 NA NA NA N N . 2 differentiated HC 147 moderately 15 NA NA NA N N 0 3 differentiated HC 148 moderately 21 NA NA NA Y N . . differentiated HC 149 NA 8 NA NA NA N N 0 0 HC 150 moderately 13 NA NA NA N . 0 3 differentiated HC 151 well 6.5 NA NA NA N Y 2 4 differentiated HC 152 well 3.5 NA NA NA N N 0 2 differentiated HC 153 well 5 NA NA NA N . 0 3 differentiated HC 154 well 45 Y Y 25 N N Y N 0 1 differentiated HC 155 well 24 N N 1 N N N Y 2 4 differentiated HC 156 moderately 70 N Y 16 Y N N Y 2 4 differentiated HC 157 well 140 Y Y 2 N N Y N 0 1 differentiated HC 159 well 35 N N NA N N N Y 2 4 differentiated HC 161 well 210 N Y 2 N N N N 1 1 differentiated HC 162 moderately 130 Y Y 77 N N Y N 0 0 differentiated HC 163 moderately 80 N Y 4 N N N Y 1 4 differentiated HC 164 well 90 N N 1 N N Y N 0 1 differentiated HC 165 well 30 N Y 4 N N N N 0 2 differentiated HC 168 well 25 N N 1 Y Y N Y 2 4 differentiated HC 169 well 35 N N NA N N N Y 2 4 differentiated HC 170 well 220 N N 0 N N Y N 0 0 differentiated HC 171 Peu 70 Y Y 10 Y N N N 1 2 différencié HC 172 moderately 40 N Y 28 N N N N 2 3 differentiated HC 173 well 40 N N 0 N N Y N 0 0 differentiated HC 176 well 75 N N NA N N Y N 0 0 differentiated HC 177 moderately 2.3 NA N NA Y A1 F4 differentiated HC 178 well 6.5 NA N NA Y A1 F4 differentiated HC 179 well-moder- 9 NA Y NA Y A2 F1 ate-poor HC 180 moderately 15 NA Y NA Y A2 F2 differentiated HC 181 well 3.5 NA Y NA Y A1 F4 moderately HC 182 well 11 NA N NA N F1 differentiated HC 183 well 8 NA Y NA N A1 F3 differentiated HC 184 well 6.5 NA N NA N F1 differentiated HC 185 moderately 3.5 NA N NA N A1 F4 differentiated HC 186 well 17 NA Y NA N F0 moderately HC 187 well 8 NA Y NA N F4 differentiated HC 188 well 13 NA N NA N F0 differentiated HC 189 well 22 NA Y NA Y F1 differentiated HC 190 moderately 15 NA N NA Y A1 F3 differentiated chronic HC 000 tumor viral Etiology Etiology identification hepatitis HBV HCV alcool Hemochromatos —NASH HC 001 N N N Y N N HC 003 Y N Y N N N HC 004 N N N N N N HC 006 N N N Y Y N HC 007 N N N Y N N HC 008 Y N Y N N N HC 009 N N N Y N N HC 010 Y Y N N N N HC 011 Y Y Y N N N HC 012 Y Y N N N N HC 014 Y N Y Y N N HC 015 N N N Y N N HC 017 Y Y N N N N HC 018 N N N Y N N HC 020 N N N Y N N HC 021 N N N Y N N HC 022 N N N Y N N HC 023 N N N N N N HC 025 N N N N N N HC 026 Y Y N N N N HC 027 Y N Y N N N HC 028 N N N N N N HC 030 N N N Y N N HC 032 Y N Y N N N HC 034 N N N N N N HC 037 N N N Y N N HC 038 Y N Y N N N HC 041 Y N Y N N N HC 042 Y Y N N N N HC 043 Y N Y N N N HC 052 Y Y N N N N HC 058 Y N Y N N N HC 060 HC 064 Y N Y N N N HC 066 Y Y N Y N N HC 101 Y Y Y Y N N HC 102 Y Y Y N N N HC 103 Y Y N N N N HC 104 N N N N N N HC 105 Y N Y N N N HC 106 Y Y N N N N HC 107 N N N Y N N HC 108 Y N Y N N N HC 109 N N N Y N Y HC 110 Y N Y Y N N HC 111 N N N Y N N HC 112 N N N N N Y HC 113 Y N Y N N N HC 114 N N N Y N N HC 115 N N N N N Y HC 116 Y Y N N N N HC 117 Y N Y N N N HC 118 N N N N N N HC 119 Y N Y Y N N HC 120 Y Y N N N N HC 121 N N N Y N Y HC 122 Y Y N N N N HC 123 Y N Y N N N HC 124 N N N N N Y HC 125 N N N Y N N HC 126 N N N N N N HC 127 Y Y N N N N HC 128 N N N Y N N HC 129 Y N Y N N N HC 130 Y Y N N N N HC 131 N N N N N N HC 137 N N N Y N N HC 138 N N N N N N HC 139 N N N N N N HC 140 N N N N N N HC 141 N N N N N N HC 142 N N N Y N N HC 143 N N N N Y N HC 144 N N N N N N HC 145 N N N Y N N HC 146 Y Y N Y N N HC 147 N N N N Y N HC 148 N N N N N N HC 149 N N N N N N HC 150 N N N Y N N HC 151 N N N N N Y HC 152 N N N N Y N HC 153 Y Y N Y N N HC 154 N N N N N N HC 155 N N N Y N N HC 156 Y N Y N N N HC 157 N N N N N N HC 159 N N N Y N N HC 161 N N N N N N HC 162 N N N N N N HC 163 N N N Y N N HC 164 N N N N N N HC 165 N N N N Y N HC 168 Y N Y N N N HC 169 N N N Y N N HC 170 N N N N N N HC 171 N N N Y N N HC 172 N N N N Y N HC 173 N N N N N N HC 176 N N N N N N HC 177 Y Y N N N N HC 178 N N N Y N N HC 179 Y N Y N N N HC 180 Y Y N N N N HC 181 N N N Y N N HC 182 Y N Y N N N HC 183 Y Y N N N N HC 184 N N N Y N N HC 185 Y N Y N N N HC 186 NA NA NA NA NA NA HC 187 N N N Y N N HC 188 N N N Y N N HC 189 N N N Y N N HC 190 Y Y N N N N

REFERENCES

  • Assou, S., Le Carrour, T., Tondeur, S., Strom, S., Gabelle, A., Marty, S., Nadal, L., Pantesco, V., Reme, T., Hugnot, J. P., et al. (2007). A meta-analysis of human embryonic stem cells transcriptome integrated into a web-based expression atlas. Stem Cells 25, 961-973.
  • Boyault, S., Rickman, D. S., de Reynies, A., Balabaud, C., Rebouissou, S., Jeannot, E., Herault, A., Saric, J., Belghiti, J., Franco, D., et al. (2007). Transcriptome classification of HCC is related to gene alterations and to new therapeutic targets. Hepatology 45, 42-52.
  • Finegold, M. J., Lopez-Terrada, D. H., Bowen, J., Washington, M. K., and Qualman, S. J. (2007). Protocol for the examination of specimens from pediatric patients with hepatoblastoma. Arch Pathol Lab Med 131, 520-529.
  • Fodde, R., and Brabletz, T. (2007). Wnt/beta-catenin signaling in cancer sternness and malignant behavior. Curr Opin Cell Biol 19, 150-158.
  • Glinsky, G. V., Berezovska, O., and Glinskii, A. B. (2005). Microarray analysis identifies a death-from-cancer signature predicting therapy failure in patients with multiple types of cancer. J Clin Invest 115, 1503-1521.
  • Hirschman, B. A., Pollock, B. H., and Tomlinson, G. E. (2005). The spectrum of APC mutations in children with hepatoblastoma from familial adenomatous polyposis kindreds. J Pediatr 147, 263-266.
  • Irizarry, R. A., Hobbs, B., Collin, F., Beazer-Barclay, Y. D., Antonellis, K. J., Scherf, U., and Speed, T. P. (2003). Exploration, normalization, and summaries of high density oligonucleotide array probe level data. Biostatistics 4, 249-264.
  • Lee, J. S., Heo, J., Libbrecht, L., Chu, I. S., Kaposi-Novak, P., Calvisi, D. F., Mikaelyan, A., Roberts, L. R., Demetris, A. J., Sun, Z., et al., (2006). A novel prognostic subtype of human hepatocellular carcinoma derived from hepatic progenitor cells. Nat Med 12, 410-416.
  • McLin, V. A., Rankin, S. A., and Zorn, A. M. (2007). Repression of Wnt/β-catenin signaling in the anterior endoderm is essential for liver and pancreas development. Development 134, 2207-2217.
  • Perilongo, G., Shafford, E., and Plaschkes, J. (2000). SIOPEL trials using preoperative chemotherapy in hepatoblastoma. Lancet Oncol 1, 94-100.
  • Rowland, J. M. (2002). Hepatoblastoma: assessment of criteria for histologic classification. Med Pediatr Oncol 39, 478-483.
  • Schnater, J. M., Kohler, S. E., Lamers, W. H., von Schweinitz, D., and Aronson, D. C. (2003). Where do we stand with hepatoblastoma? A review. Cancer 98, 668-678.
  • Taniguchi, K., Roberts, L. R., Aderca, I. N., Dong, X., Qian, C., Murphy, L. M., Nagorney, D. M., Burgart, L. J., Roche, P. C., Smith, D. I., et al. (2002). Mutational spectrum of beta-catenin, AXIN1, and AXIN2 in hepatocellular carcinomas and hepatoblastomas. Oncogene 21, 4863-4871.
  • Wei, Y., Fabre, M., Branchereau, S., Gauthier, F., Perilongo, G., and Buendia, M. A. (2000). Activation of beta-catenin in epithelial and mesenchymal hepatoblastomas. Oncogene 19, 498-504.
  • Lustgarten, J. L. et al (2008)—Improving classification performance with discretization on biomedical datasets. AMIA 2008 Symposium Proceedings,

Claims

1. Method to determine the gene expression profile on a biological sample, comprising:

a. assaying the expression of a set of genes in a sample previously obtained from a patient diagnosed for a liver tumor, wherein said set comprises from 2 to 16 genes or consists of 2 to 16 genes, said 2 to 16 genes being chosen in the group consisting in the alpha-fetoprotein (AFP), aldehyde dehydrogenase 2 (ALDH2), amyloid P component serum (APCS), apolipoprotein C-IV (APOC4), aquaporin 9 (AQP9), budding uninhibited by benzimidazoles 1 (BUB1), complement componant 1 (C1S), cytochrome p450 2E1 (CYP2E1), discs large homolog 7 (DLG7), dual specificity phosphatase 9 (DUSP9), E2F5 transcription factor (E2F5), growth hormone receptor (GHR), 4-hydroxyphenylpyruvase dioxygenase (DHP), immunoglogulin superfamily member 1 (IGSF1), Notchless homolog 1 (NLE1) and the ribosomal protein L10a (RPL10A) genes; and
b. determining the gene expression profile of said sample.

2. Method according to claim 1, which further comprises determining the grade of the liver tumor providing the sample, for example by comparing the obtained gene expression profile of said sample to the gene expression profile of a reference sample or to the gene expression profiles of a collection of reference samples or by applying a discretization method for classification.

3. Method according to claim 1 or 2, wherein the assay of the expression of said set of genes comprises a step of detecting nucleotide targets, wherein each nucleotide target is a product resulting from the expression of one of the genes in said set.

4. Method according to claim 2, wherein said nucleotide targets are mRNA.

5. Method according to any one of claims 1 to 4, wherein the assay of the expression of said set of genes comprises an amplification step, such as performed by qualitative polymerase chain reaction prior to a step of detecting the mRNA of each gene of said set.

6. Method according to any one of claims 1 to 5, wherein the assay of the expression of said set of genes comprises a hybridization step, such as one performed by hybridization on a solid or liquid support, especially on an array, prior to a step of detecting the mRNA of each gene of said set.

7. Method according to any one of claims 2 to 5, wherein said detected nucleotide targets are quantified with respect to at least one nucleotide target, expression product of an invariant gene, such as ACTG1, EFF1A1, PNN and RHOT2 genes.

8. Method according to any one of claims 1 to 7, wherein said liver tumor is a hepatoblastoma (HB) or a hepatocellular carcinoma (HCC).

9. Method according to any one of claims 2 to 8, wherein said method comprises, before step a., the preparation of said nucleotide targets from the sample.

10. Method according to any one of claims 1 to 9, wherein said set of genes comprises or consists in a set chosen in the group consisting of:

(a) E2F5 and HPD genes;
(b) APCS, BUB1, E2F5, GHR and HPD genes;
(c) ALDH2, APCS, APOC4, BUB1, C1S, CYP2E1, E2F5, GHR and HPD genes;
(d) ALDH2, APCS, APOC4, AQP9, BUB1, C1S, DUSP9, E2F5 and RPL10A genes;
(e) ALDH2, APCS, APOC4, AQP9, C1S, CYP2E1, E2F5, GHR, IGSF1 and RPL10A genes; and
(f) AFP, ALDH2, APCS, APOC4, AQP9, BUB1, C1S, CYP2E1, DLG7, DUSP9, E2F5, GHR, HPD, IGSF1, NLE1 and RPL10A genes.

11. Method enabling the determination of the tumor grade on a patient's biological sample, which comprises a classification of the tumor through discretization according the following steps:

In a method according to any of claims 1 to 10, measuring the expression and especially the relative (normalized) expression of each gene in a set of genes defined as the signature of the tumor, for example by quantitative PCR thereby obtaining data as Ct or preferably Delta Ct in said biological sample wherein said set of genes is divided in two groups, a first group consisting of the proliferation-related genes and a second group consisting of the differentiation-related genes,
comparing the values measured for each gene, to a cut-off value determined for each gene of the set of genes, and assigning a discretized value to each of said measured expression values with respect to said cut-off value, said discretized value being advantageously a “1” or a “2” and optionally a “1.5” value with respect to the cut-off value,
determining the average of the discretized values for the genes, in each group of the set of genes,
determining a score calculated as a ratio the average for the discretized values for the proliferating-related genes on the average for the discretized values for the differentiation-related genes,
comparing the obtained score for the biological sample with one or more sample cut-off(s) value(s), wherein each cut-off value corresponds to a selected percentile,
determining the tumor grade as C1 or C2, as a result of the classification of the biological sample with respect to said sample cut-off.

12. Method according to claim 11, wherein the relative expression determined for the profiled gene is obtained by normalizing with respect to the invariant RHOT2 gene.

13. Method according to claim 11, wherein the determination of the tumor grade on a biological sample comprises applying the following conditions:

a) for a hepatoblastoma:
the set of assayed genes for profiling is constituted of the 16 genes disclosed;
the invariant gene (of reference) is RHOT2;
the cut-offs value for each gene are:
AFP: 3.96139596; ALDH2: 4.3590482; APCS: 4.4691582; APOC4: 2.03068712; AQP9: 3.38391456; BUB1: −1.41294708; C1S: 4.24839464; CYP2E1: 6.70659644; DLG7: −3.3912188; DUSP9: 2.07022648; E2F5: −0.72728656; GHR: −0.1505569200; HPD: 2.27655628; IGSF1: 0.1075015200; NLE: −0.02343571999; RPL10A: 6.19723876.
the cut-off value for the sample is 0.91 and a sample with a score above 0.91 is classified into the C2 class and a sample with a score below 0.91 is classified into the C1 class.
b) for a hepatocellular carcinoma:
the set of assayed genes for profiling is constituted of the 16 genes disclosed;
the invariant gene (of reference) is RHOT2;
the cut-offs value for each gene is:
Gene name Cut-off for Taqman Cut-off for SybrGreen AEP −1.2634010 −2.3753035 ALDH2 4.014143 5.314302 APCS 5.6142907 6.399079 APOC4 −0.7963158 4.656336 AQP9 4.2836011 5.446966 BUB1 −1.2736579 −3.634476 C1S 6.3514679 6.240002 CYP2E1 6.9562419 5.829384 DLG7 −2.335694 −4.614352 DUSP9 −7.979559 −1.8626715 E2F5 −0.4400218 −1.367846 GHR 1.0832632 1.169362 HPD 6.480328 6.736329 IGSF1 −4.8417785 7.6653982 NLE −1.6167268 −1.82226 RPL10A 6.2483056 5.731897
the cut-off value for the sample corresponding to the 67th percentile is 0.925 and the cut-off value corresponding to the 33th percentile is 0.66 and a sample with a score above 0.925 is classified into the C2 class and a sample with a score below 0.66 is classified into the C1 class.

14. Method according to claim 13 wherein in the case of a hepatocellular carcinoma, a sample with a score (initial score) between 0.66 and 0.925 is refined to obtain a modified score, the modified score being either “1” or “2” depending on the calculated average of the discretized values for the proliferation-related genes only, said average being discretized at a determined percentile (the 60th for example) and “1” is assigned if the sample has an average below the value at the percentile of reference and “2” is assigned if the sample has an average above the value at the percentile of reference.

15. Kit, suitable to carry out the method as defined in any one of claims 1 to 13, comprising

a. a plurality of pairs of primers specific for a set of genes to be assayed, said set comprising 2 to 16 genes or consisting of 2 to 16 genes, said 2 to 16 genes being chosen in the group consisting of AFP, ALDH2, APCS, APOC4, AQP9, BUB1, C1S, CYP2E1, DLG7, DUSP9, E2F5, GHR, HPD, IGSF1, NLE1 and RPL10A genes; and
b. optionally reagents necessary for the amplification of the nucleotide targets of these genes by said primers, and optionally reagents for detecting the amplification products.

16. Kit according to claim 14, wherein each primer is 10 to 30 bp in length and has at least 80% similarity with its complementary sequence in the nucleotide target, preferably 100%.

17. Kit according to claim 14 or 15, wherein said pairs of primers are chosen in the group consisting of: Forward primer gene (5′-3′) Reverse primer (5′-3′) AFP AACTATTGGCCTGTGGCGAG TCATCCACCACCAAGCTGC ALDH2 GTTTGGAGCCCAGTCACCCT GGGAGGAAGCTTGCATGATTC APCS GGCCAGGAATATGAACAAGCC CTTCTCCAGCGGTGTGATCA APOC4 GGAGCTGCTGGAGACAGTGG TTTGGATTCGAGGAACCAGG AQP9 GCTTCCTCCCTGGGACTGA CAACCAAAGGGCCCACTACA BUB1 ACCCCTGAAAAAGTGATGCCT TCATCCTGTTCCAAAAATCCG C1S TTGTTTGGTTCTGTCATCCGC TGGAACACATTTCGGCAGC CYP2E1 CAACCAAGAATTTCCTGATC AAGAAACAACTCCATGCGAGC CAG DLG7 GCAGGAAGAATGTGCTGAAA TCCAAGTCTTTGAGAAGGGCC CA DUSP9 CGGAGGCCATTGAGTTCATT ACCAGGTCATAGGCATCGTTG E2F5 CCATTCAGGCACCTTCTGGT ACGGGCTTAGATGAACTCGACT GHR CTTGGCACTGGCAGGATCA AGGTGAACGGCACTTGGTG HPD ATCTTCACCAAACCGGTGCA CCATGTTGGTGAGGTTACCCC IGSF1 CACTCACACTGAAAAACGCCC GGGTGGAGCAATTGAAAGTCA NLE1 ATGTGAAGGCCCAGAAGCTG GAGAACTTCGGGCCGTCTC RPL10A TATCCCCCACATGGACATCG TGCCTTATTTAAACCTGGGCC

and,
a modified group of primers with respect to the above, wherein one or more primer(s) is modified, provided said primer(s) has at least 80% similarity with its non-modified version above.

18. A set of probes, suitable to carry out the method as defined in any one of claims 1 to 13, comprising a plurality of probes specific for a set of genes to assay, said set comprising or having from 2 to 16 genes, said 2 to 16 genes being chosen in the group consisting of AFP, ALDH2, APCS, APOC4, AQP9, BUB1, C1S, CYP2E1, DLG7, DUSP9, E2F5, GHR, HPD, IGSF1, NLE1 and RPL10A genes.

19. A set of probes according to claim 17, wherein said probes are 50 to 200 bp in length and have at least 80% similarity to the complementary sequence of the nucleotide target of the gene, preferably 100%.

20. A solid support, especially an array comprising a set of probes as defined in claims 17 or 18 linked to a support.

21. A composition comprising a set of probes as defined in claim 17 or 18, in solution.

22. A kit comprising a set of probes as defined in claim 17 or 18, a solid support as defined in claim 19 or a composition as defined in claim 20, and optionally reagents necessary for the hybridization of said nucleotide targets to said probes.

23. Set of probes, solid support, arrays, compositions or kits according to any one of claims 14 to 21, suitable for assaying a set of genes which comprises or consists in a set chosen in the group consisting of:

(a) E2F5 and HPD genes;
(b) APCS, BUB1, E2F5, GHR and HPD genes;
(c) ALDH2, APCS, APOC4, BUB1, C1S, CYP2E1, E2F5, GHR and HPD genes;
(d) ALDH2, APCS, APOC4, AQP9, BUB1, CIS, DUSP9, E2F5 and RPL10A genes;
(e) ALDH2, APCS, APOC4, AQP9, C1S, CYP2E1, E2F5, GHR, IGSF1 and RPL10A genes; and
(f) AFP, ALDH2, APCS, APOC4, AQP9, BUB1, C1S, CYP2E1, DLG7, DUSP9, E2F5, GHR, HPD, IGSF1, NLE1 and RPL10A genes.

24. Set of probes, solid support, arrays, compositions or kits according to any one of claims 14 to 23, wherein the invariant gene is the RHOT2 gene or the PNN gene.

25. Use of a set of probes, solid support, arrays, compositions or kits according to any one of claims 14 to 24, to determine the grade of a liver tumor in a sample obtained from a patient.

26. Use according to claim 25 or method of claim 11, wherein for a hepatoblastoma or for a hepatocellular carcinoma the cut-off value of the profiled genes are determined for the overexpressed proliferation-related genes at a percentile within the range of the 60th to the 80th percentile, especially at the 67th percentile and the cut-off value of the profiled genes are determined for the downregulated differentiation-related genes at a percentile within the range of the 30rd to 45th percentile, especially at the 33rd or 40th percentile and the cut-off value of the sample is determined within the same range of the 60th to the 80th percentile.

27. Use of a set of probes, arrays, compositions or kits according to any one of claims 14 to 21, to determine, in a patient, the risk of developing metastasis.

28. Use of a set of probes, arrays, compositions or kits according to any one of claims 14 to 21, to define the therapeutic regimen to apply to said patient.

Patent History
Publication number: 20120040848
Type: Application
Filed: Jun 26, 2009
Publication Date: Feb 16, 2012
Patent Grant number: 9347088
Applicants: INSTITUT PASTEUR (Paris Cedex 15), CENTRE NATIONAL DE LA RECHERCHE SCIENTIFIQUE (CNRS) (Paris Cedex 16), INSTITUT NATIONAL DE LA SANTE ET DE LA RECHERCHE MEDICALE (INSERM) (Paris Cedex 13)
Inventors: Marie-Annick Buendia (Le Perreux sur Marne), Carolina Niell (Blanes (Girona)), Stefano Cairo (Longpont-sur-Orge), Aurelien de Reynies (Paris)
Application Number: 12/999,907
Classifications
Current U.S. Class: 506/9.000; 435/6.120; 506/16.000; 435/6.100; 435/6.180
International Classification: C40B 30/04 (20060101); C40B 40/06 (20060101); C12Q 1/68 (20060101);